PLS Models (pl + models)

Distribution by Scientific Domains


Selected Abstracts


Some common misunderstandings in chemometrics

JOURNAL OF CHEMOMETRICS, Issue 7-8 2010
Karin Kjeldahl
Abstract This paper describes a number of issues and tools in practical chemometric data analysis that are often either misunderstood or misused. Deciding what are relevant samples and variables, (mis-)use of common model diagnostics, and interpretational issues are addressed in relation to component models such as PCA and PLS models. Along with simple misunderstandings, the use of chemometric software packages may contribute to the mistakes if not used critically, and it is thus a main conclusion that good data analysis practice requires the analyst to take responsibility and do what is relevant for the given purpose. Copyright © 2010 John Wiley & Sons, Ltd. [source]


Stacked partial least squares regression analysis for spectral calibration and prediction

JOURNAL OF CHEMOMETRICS, Issue 10 2009
Wangdong Ni
Abstract Two novel algorithms which employ the idea of stacked generalization or stacked regression, stacked partial least squares (SPLS) and stacked moving-window partial least squares (SMWPLS) are reported in the present paper. The new algorithms establish parallel, conventional PLS models based on all intervals of a set of spectra to take advantage of the information from the whole spectrum by incorporating parallel models in a way to emphasize intervals highly related to the target property. It is theoretically and experimentally illustrated that the predictive ability of these two stacked methods combining all subsets or intervals of the whole spectrum is never poorer than that of a PLS model based only on the best interval. These two stacking algorithms generate more parsimonious regression models with better predictive power than conventional PLS, and perform best when the spectral information is neither isolated to a single, small region, nor spread uniformly over the response. A simulation data set is employed in this work not only to demonstrate this improvement, but also to demonstrate that stacked regressions have the potential capability of predicting property information from an outlier spectrum in the prediction set. Moisture, oil, protein and starch in Cargill corn samples have been successfully predicted by these new algorithms, as well as hydroxyl number for different instruments of terpolymer samples including and excluding an outlier spectrum. Copyright © 2009 John Wiley & Sons, Ltd. [source]


Analysis of low content drug tablets by transmission near infrared spectroscopy: Selection of calibration ranges according to multivariate detection and quantitation limits of PLS models

JOURNAL OF PHARMACEUTICAL SCIENCES, Issue 12 2008
Manel Alcalà
Abstract The content uniformity of low dose products is a major concern in the development of pharmaceutical formulations. Near infrared spectroscopy may be used to support the design and optimization of potent drug manufacturing processes through the analysis of blends and tablets in a relatively short time. A strategy for the selection of concentration ranges in the development of multivariate calibration is presented, evaluating the detection and quantitation limits of the obtained multivariate models. The strategy has been applied to the determination of an active principle in pharmaceutical tablets of low concentration (0,5%, w/w), using Fourier Transform Near Infrared (FT-NIR) transmission spectroscopy. The quantitation and detection limits decreased as the upper concentration level of the calibration models was reduced. The results obtained show that the selection of concentration ranges is a critical aspect during model design. The selection of wide concentration ranges with high levels is not recommended for the determination of analytes at minor levels (<1%, w/w), even when the concentration of interest is within the range of the model. © 2008 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 97:5318,5327, 2008 [source]


CP-MLR/PLS Directed Structure-Activity Modeling of the HIV-1 RT Inhibitory Activity of 2,3-Diaryl-1,3-thiazolidin-4-ones

MOLECULAR INFORMATICS, Issue 4 2004
Yenamandra
Abstract A detailed structure-activity relationship study of the HIV-1 Reverse Transcriptase (RT) inhibitory activity of two series of compounds, 2-(2,6-dihalo phenyl)-3-(substituted pyridin-2-yl)-thiazolidin-4-ones and 2-(2,6-Dihalophenyl)-3-(substituted phenyl)-thiazolidin-4-ones, belonging to 2,3-diaryl-thiazolidin-4-ones in terms of physicochemical and structural descriptors have been carried out using combinatorial protocol interfaced multiple linear regression (CP-MLR) and partial least squares (PLS) analysis. The models developed in the study indicate a preference for hydrophobic compounds for better inhibitory activity. Also, a positive regression coefficient of I2,3, an indicator descriptor meant for the joint disposition of substituents of 2,3-diaryl moieties of thiazolidinones to address their ability in taking a butterfly like conformation, is in agreement with all earlier observations that compounds having the ability to take butterfly like conformation will be showing better inhibitory activity. The identified models suggest that the meta-positions of 3-aryl moiety has the ability to accommodate hydrophobic/ steric groups. The replacement of C2, of 3-phenyl by nitrogen resulted in 3-pyridyl variants of these analogues. Even though from geometry point of view, the phenyls and pyridyls span almost the same structural space and steric features, presence of nitrogen in pyridyls gave them a blend of polarity, electronic features and a different hydrophobicity profile when compared to corresponding phenyl analogue. Furthermore the PLS models, developed from those descriptors took part in CP-MLR models, indicate that most of the descriptors have almost equal influence in accounting for the variation in the activity of these compounds. This suggests that the factors responsible for the variation in the activity have been uniformly distributed across the varying centers of the molecule. The study suggests that thiazolidinones with 3-(pyridin-2-yl) moiety provide scope for further substituent variation to modulate the HIV-1 RT inhibitory activity. [source]


Local dynamic partial least squares approaches for the modelling of batch processes

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, Issue 5 2008
N. M. Fletcher
Abstract The application of multivariate statistical projection based techniques has been recognized as one approach to contributing to an increased understanding of process behaviour. The key methodologies have included multi-way principal component analysis (PCA), multi-way partial least squares (PLS) and batch observation level analysis. Batch processes typically exhibit nonlinear, time variant behaviour and these characteristics challenge the aforementioned techniques. To address these challenges, dynamic PLS has been proposed to capture the process dynamics. Likewise approaches to removing the process nonlinearities have included the removal of the mean trajectory and the application of nonlinear PLS. An alternative approach is described whereby the batch trajectories are sub-divided into operating regions with a linear/linear dynamic model being fitted to each region. These individual models are spliced together to provide an overall nonlinear global model. Such a structure provides the potential for an alternative approach to batch process performance monitoring. In the paper a number of techniques are considered for developing the local model, including multi-way PLS and dynamic multi-way PLS. Utilising the most promising set of results from a simulation study of a batch process, the local model comprising individual linear dynamic PLS models was benchmarked against global nonlinear dynamic PLS using data from an industrial batch fermentation process. In conclusion the results for the local operating region techniques were comparable to the global model in terms of the residual sum of squares but for the global model structure was evident in the residuals. Consequently, the local modelling approach is statistically more robust. L'application de techniques basées sur la projection statistique multivariée est reconnue comme étant une approche qui contribue à une meilleure compréhension du comportement des procédés. Les méthodologies clés incluent l'analyse des composantes principales (PCA) à plusieurs critères de classification, les moindres carrés partiels (PLS) à plusieurs critères de classification et l'analyse des niveaux d'observation discontinus. Les procédés discontinus présentent typiquement un comportement non linéaire et variable dans le temps et ces caractéristiques mettent au défi les techniques mentionnées ci-dessus. Devant ces défis, la méthode PLS dynamique est proposée pour saisir la dynamique des procédés. Des approches semblables pour supprimer la non linéarité des procédés incluent le retrait de la trajectoire principale et l'application des PLS non linéaires. On décrit une autre approche où les trajectoires discontinues sont subdivisées en régions opératoires avec un modèle dynamique linéaire/linéaire adapté à chaque région. Ces modèles individuels sont raccordés pour obtenir un modèle non linéaire global. Une telle structure présente un potentiel pour une approche différente du suivi des performances des procédés discontinus. Dans cet article, plusieurs techniques sont considérées pour la mise au point du modèle local, incluant les PLS à plusieurs critères de classification et les PLS à plusieurs critères de classification dynamique. En utilisant la série de résultats les plus prometteurs d'une étude de simulation d'un procédé discontinu, le modèle local comprenant les modèles de PLS dynamiques linéaires individuels a été comparé à la méthode de PLS non linéaires dynamique globale utilisant des données d'un procédé de fermentation discontinu industriel. En conclusion, les résultats pour les techniques des régions opératoires locales sont comparables au modèle global en termes de somme des carrés des résidus mais pour le modèle global, la présence d'une structure dans les résidus est évidente. En conséquence, l'approche de modélisation locale est statistiquement plus robuste. [source]