Multivariate Partial Least-squares Research Articles

Performance Prediction of Waterflooding in Western Canadian Heavy Oil Reservoirs Using Artificial Neural Network

This study presents a successful application of multivariate partial least-squares (PLS), response surface methodology (RSM), and artificial neural network (ANN) to develop a new diagnostic tool for performance prediction of waterflooding in heavy oil reservoirs. The data used in this study consist of 120 operational and reservoir parameters for 177 waterfloods in heavy and medium oil reservoirs in western Canada (i.e., Alberta and Saskatchewan). This study also used 15 numerically devised indices for performance evaluation of water injection, based on collected injection and production histories of studied waterfloods. To reduce the number and complexity of input parameters of ANN models, a comprehensive PLS analysis was conducted. In addition to parameter selection, the PLS model provided a more in-depth understanding of differences between heavy and medium oil waterfloods. Next, RSM was used to improve the quality of the database for a selected combination of 38 reservoir and operational parameters by predicting some of the missing data. Finally, ANN models were developed using the feed-forward backpropagation algorithm with momentum for error minimization. The developed models show the superior ability of the ANN for creation of an efficient reservoir engineering tool for fast performance prediction of waterfloods using easily obtainable operational and reservoir parameters. The developed models in this study can be incorporated into reservoir engineering, risk assessment, and production optimization software programs to improve the quality of predictions based on more than 50 years of waterflooding experience in western Canada.

Evaluation of the potential of Raman microspectroscopy for prediction of chemotherapeutic response to cisplatin in lung adenocarcinoma

The study of the interaction of anticancer drugs with mammalian cells in vitro is important to elucidate the mechanisms of action of the drug on its biological targets. In this context, Raman spectroscopy is a potential candidate for high throughput, non-invasive analysis. To explore this potential, the interaction of cis-diamminedichloroplatinum(II) (cisplatin) with a human lung adenocarcinoma cell line (A549) was investigated using Raman microspectroscopy. The results were correlated with parallel measurements from the MTT cytotoxicity assay, which yielded an IC(50) value of 1.2 ± 0.2 µM. To further confirm the spectral results, Raman spectra were also acquired from DNA extracted from A549 cells exposed to cisplatin and from unexposed controls. Partial least squares (PLS) multivariate regression and PLS Jackknifing were employed to highlight spectral regions which varied in a statistically significant manner with exposure to cisplatin and with the resultant changes in cellular physiology measured by the MTT assay. The results demonstrate the potential of the cellular Raman spectrum to non-invasively elucidate spectral changes that have their origin either in the biochemical interaction of external agents with the cell or its physiological response, allowing the prediction of the cellular response and the identification of the origin of the chemotherapeutic response at a molecular level in the cell.

Predictive factors for disability outcome at twenty weeks and two years following a pain self-management group intervention in patients with persistent neck pain in primary health care

To explore possible predictors associated with short-term (post-treatment) and long-term (2 years) treatment success in terms of pain-related disability for patients with persistent neck pain following a pain and stress self-management intervention (PASS). Data from 77 participants assigned to PASS in a randomized controlled trial were explored to identify possible predictors of favourable outcome regarding pain-related disability as measured by the Neck Disability Index (NDI), by use of Pearson correlation analysis, partial least squares (PLS) and ordinary least squares (OLS) regression analyses. Data from self-assessment questionnaires completed by the participants before, post-treatment (i.e. 20 weeks after inclusion) and 2 years after inclusion in the study, were used. Multivariate PLS regression analysis showed that baseline scores in NDI, the Self-Efficacy Scale (SES) and pain intensity explained 31% of the variance in disability (NDI) post-treatment. Multivariate PLS regression analysis showed that post-treatment scores in NDI, SES and pain intensity explained 68% of the variance in disability (NDI) at 2 years. Treatment gains, as measured by post-treatment scores at 20-week follow-up, in disability, self-efficacy and pain intensity were associated with long-term outcome in pain-related disability at 2 years, in patients with persistent neck pain participating in a self-management group intervention in primary health care.

Use of Reflectance Infrared Spectroscopy for Monitoring the Metal Content of the Estuarine Sediments of the Nerbioi-Ibaizabal River (Metropolitan Bilbao, Bay of Biscay, Basque Country)

Multivariate partial least-squares (PLS) calibration models have been developed for the spatial and seasonal simultaneous monitoring of 14 trace elements (Al, As, Cd, Co, Cr, Cu, Fe, Mg, Mn, Ni, Pb, Sn, V, and Zn) in sediments from 117 samples taken in the estuary of the Nerbioi-Ibaizabal River. Models were based on the chemometric treatment of diffuse reflectance near-infrared (NIR) and attenuated total reflectance (ATR) mid infrared (MIR) spectra, obtained from samples previously lyophilized and sieved with a particle size lower than 63 microm. Vibrational spectra were scanned in both, NIR and MIR regions. Developed PLS models, based on the interaction between trace elements and organic mater provide good screening tools for the prediction of trace elements concentration in sediments.

Performance of PLS regression coefficients in selecting variables for each response of a multivariate PLS for omics-type data

Multivariate partial least square (PLS) regression allows the modeling of complex biological events, by considering different factors at the same time. It is unaffected by data collinearity, representing a valuable method for modeling high-dimensional biological data (as derived from genomics, proteomics and peptidomics). In presence of multiple responses, it is of particular interest how to appropriately "dissect" the model, to reveal the importance of single attributes with regard to individual responses (for example, variable selection). In this paper, performances of multivariate PLS regression coefficients, in selecting relevant predictors for different responses in omics-type of data, were investigated by means of a receiver operating characteristic (ROC) analysis. For this purpose, simulated data, mimicking the covariance structures of microarray and liquid chromatography mass spectrometric data, were used to generate matrices of predictors and responses. The relevant predictors were set a priori. The influences of noise, the source of data with different covariance structure and the size of relevant predictors were investigated. Results demonstrate the applicability of PLS regression coefficients in selecting variables for each response of a multivariate PLS, in omics-type of data. Comparisons with other feature selection methods, such as variable importance in the projection scores, principal component regression, and least absolute shrinkage and selection operator regression were also provided.

Linking functional and structural brain images with multivariate network analyses: A novel application of the partial least square method

In this article, we introduce a multimodal multivariate network analysis to characterize the linkage between the patterns of information from the same individual's complementary brain images, and illustrate its potential by showing its ability to distinguish older from younger adults with greater power than several previously established methods. Our proposed method uses measurements from every brain voxel in each person's complementary co-registered images and uses the partial least square (PLS) algorithm to form a combined latent variable that maximizes the covariance among all of the combined variables. It represents a new way to calculate the singular value decomposition from the high-dimensional covariance matrix in a computationally feasible way. Analyzing fluorodeoxyglucose positron emission tomography (PET) and volumetric magnetic resonance imaging (MRI) images, this method distinguished 14 older adults from 15 younger adults (p=4e−12) with no overlap between groups, no need to correct for multiple comparisons, and greater power than the univariate Statistical Parametric Mapping (SPM), multimodal SPM or multivariate PLS analysis of either imaging modality alone. This technique has the potential to link patterns of information among any number of complementary images from an individual, to use other kinds of complementary complex datasets besides brain images, and to characterize individual state- or trait-dependent brain patterns in a more powerful way.

Improving the analysis of designed studies by combining statistical modelling with study design information

BackgroundIn the fields of life sciences, so-called designed studies are used for studying complex biological systems. The data derived from these studies comply with a study design aimed at generating relevant information while diminishing unwanted variation (noise). Knowledge about the study design can be used to decompose the total data into data blocks that are associated with specific effects. Subsequent statistical analysis can be improved by this decomposition if these are applied on selected combinations of effects.ResultsThe benefit of this approach was demonstrated with an analysis that combines multivariate PLS (Partial Least Squares) regression with data decomposition from ANOVA (Analysis of Variance): ANOVA-PLS. As a case, a nutritional intervention study is used on Apoliprotein E3-Leiden (APOE3Leiden) transgenic mice to study the relation between liver lipidomics and a plasma inflammation marker, Serum Amyloid A. The ANOVA-PLS performance was compared to PLS regression on the non-decomposed data with respect to the quality of the modelled relation, model reliability, and interpretability.ConclusionIt was shown that ANOVA-PLS leads to a better statistical model that is more reliable and better interpretable compared to standard PLS analysis. From a following biological interpretation, more relevant metabolites were derived from the model. The concept of combining data composition with a subsequent statistical analysis, as in ANOVA-PLS, is however not limited to PLS regression in metabolomics but can be applied for many statistical methods and many different types of data.

The Relationship of Employee Perceptions of Organizational Climate to Business-Unit Outcomes

There has been an extensive exploration of how organizational climate is related to various business outcomes, but these studies have generally examined outcomes separately or developed univariate measures that combine outcomes. These approaches fail to (a) accommodate the multivariate character of important business results and (b) facilitate the firm's need to achieve success on several dimensions. This research proposes a methodological approach new to the service domain to address these issues. Using data from a large, multinational retail grocery superstore based in continental Western Europe, this study illustrates how multivariate partial least squares (MPLS) models can be used. MPLS provides three interpretable factors of climate—Overall Organizational Climate, Self-Efficacy Versus Leader's Efficacy, and Personal Empowerment Versus Management Facilitation —that are important predictors of three business outcomes: employee retention, customer satisfaction, and scaled revenue. The use of the MPLS approach in other services domains is also explored.

Determination of Ash Content in Solid Cattle Manure with Visible Near-Infrared Diffuse Reflectance Spectroscopy

Visible and near-infrared (VisNIR, 350-2500 nm) diffuse reflectance spectroscopy (DRS) is increasingly being used to quantify constituents of organic matter both in the lab and in situ. However, it is unknown if DRS can be utilized as a tool for determining crude ash content of solid cattle manure. Ash content is a significant contributor to the suitability and value of manure for use both as a biofuel and soil fertilizer, but conventional ash analysis is time-consuming and labor-intensive. In this study, we explored the feasibility of VisNIR-DRS for the rapid prediction of ash content in solid manure from beef feedyards in the southern High Plains. Proportionally mixed samples of soil and manure (n = 201) were evaluated for ash content by conventional analysis and then used to calibrate a statistical model for prediction of ash content by VisNIR-DRS based on multivariate partial-least squares regression and random test-set validation. Two thirds of the samples were randomly selected to build a calibration model, and the remaining third was used for validation. The coefficient of determination (r2), root mean squared deviation (RMSD), and ratio of prediction to standard deviation (RPD) were calculated to assess the prediction model. The prediction model had an r2 of 0.94, an RMSD of 5% ash (d.b.), and an RPD of 4. The VisNIR-DRS model successfully predicted crude ash content within 5% of the observed ash content (d.b.) as determined by dry oxidation using the accepted ASTM standard E1755-01.

Mapping unpleasantness of sounds to their auditory representation

Certain sounds, for example, the squeal of chalk on a blackboard, are typically perceived as highly unpleasant. This study addressed the question of what aspects of the auditory representation of such sounds are associated with judgments of unpleasantness. Participants rated the perceived unpleasantness of a large number of sounds that included "griding" and other unpleasant sounds. A multivariate partial least-squares (PLS) model was then built to relate the ratings of unpleasantness with an auditory representation derived from a model of processing in the auditory pathway. The "existence region" of unpleasantness in the auditory space of frequency-temporal modulation was determined after the PLS model had been validated by predicting the unpleasantness of novel sounds from the auditory representation. It was observed that the existence region corresponded to spectral frequencies between 2500 and 5500 Hz, and temporal modulations in the range 1-16 Hz.

Real-time on-line blend uniformity monitoring using near-infrared reflectance spectrometry: A noninvasive off-line calibration approach

A robust, noninvasive, real-time, on-line near-infrared (NIR) quantitative method is described for blend uniformity monitoring of a pharmaceutical solid dosage form containing 29.4% (w/w) drug load with three major excipients (crospovidone, lactose, and microcrystalline cellulose). A set of 21 off-line, static calibration samples were used to develop a multivariate partial least-squares (PLS) calibration model for on-line prediction of the API content during the blending process. The concentrations of the API and the three major excipients were varied randomly to minimize correlations between the components. A micro electrical-mechanical system (MEMS) based portable, battery operated NIR spectrometer was used for this study. To minimize spectral differences between the static and dynamic measurement modes, the acquired NIR spectra were preprocessed using standard normal variate (SNV) followed by second derivative Savitzky-Golay using 21 points. The performance of the off-line PLS calibration model were evaluated in real-time on 16 laboratory scale (30 L bin size) blend experiments conducted over 3 months. To challenge the robustness of the off-line calibration model, several blend experiments were conducted using a different bin size, faster revolution speed and variations in the potency of the API. Employing the PLS calibration model developed using the off-line calibration approach, the real-time API NIR (%) predictions for all experiments were all within 90–110%. These results were confirmed using the conventional thief sampling of the final blend followed by high performance liquid chromatography (HPLC) analysis. Further confirmation was established through content uniformity by HPLC of manufactured tablets. Finally, the optimized off-line PLS method was successfully transferred to a production site which involved using a secondary NIR instrument with a 15-fold scale-up in bin size from development.

Parafac and PLS Applied to Determination of Captopril in Pharmaceutical Preparation and Biological Fluids by Ultraviolet Spectrophotometry

A new ultraviolet spectrophotometric method has been developed for the direct qualitative determination of captopril in pharmaceutical preparation and biological fluids such as human plasma and urine samples. The method was accomplished based on parallel factor analysis (PARAFAC) and partial least squares (PLS). The study was carried out in the pH range from 2.0 to 12.8 and with a concentration from 0.70 to 61.50 microg ml(-1) of captopril. Multivariate calibration models PLS at various pH and PARAFAC were elaborated from ultraviolet spectra deconvolution and captopril determination. The best models for this system were obtained with PARAFAC and PLS at pH = 2.04 (PLS-PH2). The applications of the method for the determination of real samples were evaluated by analysis of captopril in pharmaceutical preparations and biological (human plasma and urine) fluids with satisfactory results. The accuracy of the method, evaluated through the root mean square error of prediction (RMSEP), was 0.58 for captopril with PARAFAC and 0.67 for captopril with PLS-PH2 model. Acidity constant of captopril at 25 degrees C and ionic strength of 0.1 M have also been determined spectrophotometrically. The obtained pKa values of captopril are 3.90 +/- 0.05 and 10.03 +/- 0.08 for pKa. and pKa2, respectively.

Quantitative determination of ethanol in heated plumes by passive Fourier transform infrared remote sensing measurements

Quantitative calibration models are developed for passive Fourier transform infrared (FT-IR) remote sensing measurements of open-air-generated vapors of ethanol. These experiments serve as a feasibility study for the use of passive FT-IR measurements in quantitative determinations of industrial stack emissions. A controlled-temperature plume generator is used to produce plumes of known concentrations of pure ethanol and mixtures of ethanol and methanol. Analyte plumes are generated over the path-averaged concentration range of 20-300 ppm-m and stack temperatures of 125, 150, 175, and 200 degrees C. A novel experimental setup is employed in which an ambient temperature polyvinyl chloride backdrop is placed behind the emission stack and used as a target for the passive IR measurements. An emission FT-IR spectrometer with telescope entrance optics is then employed to view the generated plumes against the backdrop. Signal processing techniques based on signal averaging and bandpass digital filtering are applied to both interferogram and single-beam spectral data obtained from these measurements, and the resulting filtered signals are used as inputs into the generation of multivariate partial least-squares (PLS) calibration models. Successful calibration models are obtained with both interferogram and spectral data, and neither analysis requires the collection of separate IR background data. For a set of validation data collected on a different day from the calibration measurements, standard errors of prediction of 30.6 and 32.2 ppm-m ethanol are obtained for the PLS models based on interferogram and spectral data, respectively.

Comparison of the usability of different spectral ranges within the near ultraviolet, visible and near infrared ranges (UV-VIS-NIR) region for the determination of the content of scab-damaged component in blended samples of ground wheat

Soft wheat grain samples of the same variety were obtained from a plot where the crop grew under natural conditions (control material) and from a plot where the crop was inoculated with Fusarium culmorum. The grain was ground and sieved with the finest fraction (a particle size less than 0.18 mm) of both materials being used for the preparation of samples in which the content of damaged constituent varied from zero to approximately 84%. Diffuse reflectance spectra of the absorbance from the blended samples were recorded in the 200–2500 nm spectral range and multivariate calibration PLS (Partial Least Squares) models were built within three spectral ranges: 200–2500, 200–1400 and 1400–2500 nm. Before modelling, several variants for spectra pre-processing were tried: multiple scatter correction, single and double differentiation, in all cases with and without centring. Single differentiation followed by centring was found to be the best method for spectra pre-processing in all spectral ranges. Very good calibration models were obtained for the whole and shorter wavelengths spectral ranges, allowing the detection of 1.50 and 0.76% of the content of scab-damaged constituent, respectively. Two-dimensional correlation spectroscopy applied to the set of spectra enabled the assignment of spectral bands and an analysis of changes in the chemical composition caused by scab damage. It was found that the content of protein and lipids increased with an increase of the scab-damaged constituent, whereas the content of moisture and starch decreased.

Determination of selectivity differences for basic compounds in gradient reverse phase high performance liquid chromatography under high pH conditions by partial least squares modelling

The retention behaviour of compounds in a chromatographic system is believed to be multivariate by nature, i.e. many physico-chemical properties of an analyte can influence its retention. Principal component analysis (PCA) and partial least squares (PLS) can therefore be particularly useful tools for visualising, exploring and modelling the complex interactions between solutes and the mobile and stationary phase. PCA allows the relationships between compounds (the observations) and their retention parameters (the variables) to be visualised in usually just two or three dimensions. PLS can be used to model quantitative structure-retention relationships (QSRRs) and may lead to better understanding of retention and selectivity changes in chromatographic systems. The objective of the study was to investigate the chromatographic behaviour of basic compounds under optimised gradient conditions using octadecyl high performance liquid chromatography (HPLC) columns designed for high pH separations. Three pharmaceutical mixtures were analysed by linear gradient reverse phase HPLC (RP-HPLC) at high pH using ammonia as a pH modifier, and methanol and/or acetonitrile as the organic modifier. The separations were carried out on three octadecyl columns: Waters XTerra MS C18, Agilent Zorbax Extend C18 and Thermo Hypersil-Keystone BetaBasic-18. Multivariate PCA and PLS modelling were employed to explore and explain the differences in selectivity between the chromatographic systems studied when the basic compounds were analysed under the high pH conditions. The interactions between the analytes and the mobile–stationary phases were described by relating the compound molecular descriptors with the selectivity of each chromatographic system. The selectivity differences between the chromatographic systems were identified.

Simultaneous spectrofluorimetric and spectrophotometric determination of melatonin and pyridoxine in pharmaceutical preparations by multivariate calibration methods

Partial least-squares (PLS) calibration and principal component regression (PCR) methods were utilized for the simultaneous spectrofluorimetric and spectrophotometric determination of pyridoxine (PY) and melatonin (MT). Since emission and adsorption spectra of these drugs overlap, PY and MT cannot be directly determined by fluorimetric nor by spectrophotometric methods. Full-spectrum multivariate calibration PLS and PCR methods were developed for both fluorimetry and spectrophotometry. The conditions were optimized for fluorimetric as well as for spectrophotometric determination of both drugs. The simultaneous determination of PY and MT was carried out in mixtures by recording the emission fluorescence spectrum between 324 and 500 nm ( λ ex 285 nm) for fluorimetry, and by recording the absorption spectrum between 250 and 350 nm for spectrophotometry ( λ max(PY) 310 nm, λ max(MT) 278 nm). The experimental calibration matrixes were designed orthogonally. At the optimum conditions, dynamic ranges were 0.04–1.3 and 0.1–4 μg ml –1 for fluorimetry and 1–22 and 1–24 μg ml –1 for spectrophotometry for MT and PY, respectively. The calibration concentrations were prepared in the dynamic ranges. The parameters of the chemometrics procedure for the simultaneous determination of MT and PY were optimized, and the proposed methods were validated with prediction set. Finally the procedures were successfully applied to simultaneous spectrofluorimetric and spectrophotometric determination of PY and MT in synthetic mixtures and in a pharmaceutical formulation.

Nanoliter Serum Sample Analysis by Mid-Infrared Spectroscopy for Minimally Invasive Blood-Glucose Monitoring

The aim of this study was to demonstrate that mid-infrared spectroscopy is able to quantify glucose in a serum matrix with sample volumes well below 1 muL. For this, we applied mid-infrared attenuated total reflectance (ATR) or transmission-based spectroscopic methods to glucose quantification in microsamples of dry-film sera, either undiluted or diluted 10 times in distilled water. The sample series spanned physiological glucose concentrations between 50 and 600 mg/dL and volumes of 80, 8, and 1 nL. Calibration was carried out using multivariate partial least-squares (PLS) modeling with spectral data between 1180 and 940 cm(-1). Best performance was achieved in the ATR experiments. For raw ATR spectra, the optimum standard error of prediction (SEP) of 13.3 mg/dL was obtained for the 8 nL sample series with subsequent 10-fold dilution. With respect to the coefficient of variation of the glucose assay, CV(pred), we obtained a value of 3% for the 80 nL volume samples with spectral preprocessing using matrix protein absorption bands as an internal standard, 4% for the 8 nL samples, and 6% for the 1 nL samples with raw data. Spectral standardization resulted in significant improvement, especially for the 80 nL volume sample series. By contrast, the accuracy of the glucose assay for the 1 nL sample volume series could not be improved either by internal standardization or by considering the dry film areas for normalization, which we attribute to varying topographies of the dry films.

Troubleshooting of an Industrial Batch Process Using Multivariate Methods

Multivariate statistical methods are used to analyze data from an industrial batch drying process. The objective of the study was to uncover possible reasons for major problems occurring in the quality of the product produced in the process. Partial least-squares (PLS) methods were able to isolate which group of variables in the chemistry, in the timing of the various stages of the batch, and in the shape of the time-varying trajectories of the process variables were related to a poor-quality product. The industrial study illustrates the approach and the power of these multivariate methods for troubleshooting problems occurring in complex batch processes. Several new variations in the multivariate PLS methodology for the analysis of batch data are also implemented. In particular, an application utilizing a novel approach to the time warping of the trajectories for batches, and the subsequent use of the time-warping information, is presented. The use of the time history of the PLS weights of the process variable trajectories to diagnose problems in the dynamic operation of the batches is also clearly illustrated, as is the use of contribution plots for finding features which distinguished between the operating histories of good and bad batches.

Spectrofluorimetric determination of phenylephrine in the presence of a large excess of paracetamol

Two independent spectrofluorometric methods were developed for the determination of phenylephrine (PHE) in pharmaceutical tablets containing a large excess of paracetamol (PAR) (between 30:1 and 100:1 in weight). One of them involves standard addition of phenylephrine (in the range 0–2.00 mg l −1 in excess with respect to the content of the unknown sample) to a slightly acid aqueous solution (HCl) of the tablet components. The limits of determination and quantification were found to be 0.08 and 0.27 mg l −1, respectively. A surface response optimisation approach, using a blocked cube star design was carried out to optimise the variables which have an influence on the fluorescence emission of the analyte. The optimal conditions obtained were excitation wavelength, 277 nm and concentration of paracetamol, 72 mg l −1. As an alternative method, the multivariate partial least-squares (PLS) calibration of fluorescence excitation spectra was carried out. In this case the range of application is 0.80–2.00 mg l −1 of phenylephrine. Both techniques were satisfactorily applied to several pharmaceutical tablets. The results provided by both methods are not statistically different.

Utilization of Multivariate Calibration Methods for the Study of K+ Transport through Hydrophobic Liquid Membranes by Using Isomeric Anions

This work describes utilization of the multivariate calibration method PLS (partial least squares) in the investigation of K+ transport through a choloroform membrane by using mixtures of three isomeric organic anions as counterions. A comparison between the results obtained by this method and those obtained by the stepwise multiple linear regression method was also made. The BLM (bulk liquid membrane) method was employed in the transport by using a neutral crown ether, 18-crown-6 (18C6) as the carrier. Multivariate calibration was necessary since there are superpositions among the UV-Vis spectra of the three isomers. Univariate methods are shown to be unable to quantify samples in these conditions.