Chemometrics and near-infrared spectroscopy: Avoiding the pitfalls

The use of chemometrics in quantitative near-infrared (NIR) spectroscopy is reviewed from the standpoint of avoiding pitfalls that may lead to misleading or overly optimistic results. Using the NIR analysis of glucose in six-component mixture samples as an example, a set of guidelines is presented to help the analyst develop and implement a successful calibration.     

Comparison of multivariate calibration models for glucose,urea, and lactate from near-infrared and Raman spectra

Partial least-squares (PLS) calibration models have been generated from a series of near-infrared (near-IR) and Raman spectra acquired separately from sixty different mixed solutions of glucose, lactate, and urea in aqueous phosphate buffer. Independent PLS models were prepared and compared for glucose, lactate, and urea. Near-IR and Raman spectral features differed substantially for these solutes, with Raman spectra enabling greater distinction with less spectral overlap than features in the near-IR spectra. Despite this, PLS models derived from near-IR spectra outperformed those from Raman spectra. Standard errors of prediction were 0.24, 0.11, and 0.14 mmol L⁻¹ for glucose, lactate, and urea, respectively, from near-IR spectra and 0.40, 0.42, and 0.36 mmol L⁻¹ for glucose, lactate, and urea, respectively, from Raman spectra. Differences between instrumental signal-to-noise ratios were responsible for the better performance of the near-IR models. The chemical basis of model selectivity was examined for each model by using a pure component selectivity analysis combined with analysis of the net analyte signal for each solute. This selectivity analysis showed that models based on either near-IR or Raman spectra had excellent selectivity for the targeted analyte. The net analyte signal analysis also revealed that analytical sensitivity was higher for the models generated from near-IR spectra. This is consistent with the lower standard errors of prediction.     

New cut-off criterion for uninformative variable elimination in multivariate calibration of near-infrared spectra for the determination of heroin in illicit street drugs

A new cut-off criterion has been proposed for the selection of uninformative variables prior to chemometric partial least squares (PLS) modelling. After variable elimination, PLS regressions were made and assessed comparing the results with those obtained by PLS models based on the full spectral range. To assess the prediction capabilities, uninformative variable elimination (UVE)-PLS and PLS were applied to diffuse reflectance near-infrared spectra of heroin samples. The application of the proposed new cut-off criterion, based on the t-Students distribution, provided similar predictive capabilities of the PLS models than those obtained using the original criteria based on quantile value. However, the repeatability of the number of selected variables was improved significantly.     

Quantitative analysis of the hydration of lithium salts in water using multivariate curve resolution of near-infrared spectra

The hydration process of lithium iodide, lithium bromide, lithium chloride and lithium nitrate in water was analyzed quantitatively by applying multivariate curve resolution alternating least squares (MCR-ALS) to their near infrared spectra recorded between 850 nm and 1100 nm. The experiments were carried out using solutions with a salt mass fraction between 0% and 72% for lithium bromide, between 0% and 67% for lithium nitrate and between 0% and 62% for lithium chloride and lithium iodide at 323.15 K, 333.15 K, 343.15 K and 353.15 K, respectively. Three factors were determined for lithium bromide and lithium iodide and two factors for the lithium chloride and lithium nitrate by singular value decomposition (SVD) of their spectral data matrices. These factors are associated with various chemical environments in which there are aqueous clusters containing the ions of the salts and non-coordinated water molecules. Spectra and concentration profiles of non-coordinated water and cluster aqueous were retrieved by MCR-ALS. The amount of water involved in the process of hydration of the various salts was quantified. The results show that the water absorption capacity increases in the following order LiI < LiBr < LiNO₃ < LiCl. The salt concentration at which there is no free water in the medium was calculated at each one of the temperatures considered. The values ranged between 62.6 and 65.1% for LiBr, 45.5–48.3% for LiCl, 60.4–61.2% for LiI and 60.3–63.7% for LiNO₃. These values are an initial approach to determining the concentration as from which crystal formation is favored.     

Direct and simultaneous spectrofluorometric determination of naproxen and salicylate in human serum assisted by chemometric analysis

A direct method for the simultaneous determination of naproxen and salicylate in human serum is reported, based on a combination of spectrofluorometric measurements with two multivariate calibration techniques: partial least-squares (PLS-1) and the novel net analyte preprocessing (NAP). The method is rapid, selective and sensitive, and is based on the measurement of the fluorescence spectra of NH₃ alkalinized whole human sera at the excitation wavelength of 315 nm. It can be applied within the ranges of concentrations 50-200 ng ml⁻¹ for naproxen and 100-300 ng ml⁻¹ for salicylate. The employed chemometric techniques have been compared on the basis of the statistical indicators for calibration and validation. Reproducibility and interference studies in abnormal sera have also been carried out.     

Estimation of the biological variation of glucose-6-phosphate dehydrogenase in dried blood spots

Models based on biological variation provide a well-accepted database with reliable information for clinical laboratories for all purposes including screening, diagnosis and follow-up. Newborn screening laboratories use a blood spotted paper matrix to measure the analytes. This matrix medium is certainly different than body fluid matrix medium. After long-term monitoring of the performance of the Glucose-6-Phosphate Dehydrogenase Kit (R&D Diagnostics OSMMR 2000-D G6PD), the results obtained from the variation analysis were statistically evaluated. Analytical coefficient of variation, CV _A, was found to be 5.41%. The CV _A derived from between run assays was 5.32%, while within-subject biological coefficient of variation, CV _I, was 7.26%. Since minimum performance is defined as CV _A< 0.750 CV _I , CV _A should be lower than 5.44%. The analytical bias in calculation of total error was chosen to evaluate the performance of this assay. In this aspect, CV _G, between-subject biological coefficient of variation, was found to be equal to 10.35%. B _A was found to be 4.12%, which is lower than 4.74%, which means that it is acceptable. Therefore, the minimum quality specification for total error allowable is . When the relevant results obtained in this study were substituted in this formula, TE _a was found to be 13.7% for G6PD measurement in dried blood spots on paper filter matrix. It is expected that this figure will be helpful for the performance evaluation of newborn screening laboratories performing G6PD screening. We have been using error grid graphs for the evaluation of our external quality assurance survey results for the last two years, only because there was no data for assays employing filter matrix. Even the TE _a already reported for EDTA whole blood samples used in G6PD assays has been remarkably high, which can easily create the wrong impression that G6PD is not a reliable test to perform from blood spot cards. The present study shows that this assay is adequately reliable even when performed from dried blood spot matrix. However, we believe that the combination of total allowable error, error grid graphs with a well-defined cut off is the best approach to obtain an accurate performance evaluation for this test.Presented at the 10th Conference Quality in the Spotlight, March 2005, Antwerp, Belgium.     

Multivariate analysis of remote laser-induced breakdown spectroscopy spectra using partial least squares,principal component analysis,and related techniques

Quantitative analysis with laser-induced breakdown spectroscopy traditionally employs calibration curves that are complicated by chemical matrix effects. These chemical matrix effects influence the laser-induced breakdown spectroscopy plasma and the ratio of elemental composition to elemental emission line intensity. Consequently, laser-induced breakdown spectroscopy calibration typically requires a priori knowledge of the unknown, in order for a series of calibration standards similar to the unknown to be employed. In this paper, three new Multivariate Analysis techniques are employed to analyze the laser-induced breakdown spectroscopy spectra of 18 disparate igneous and highly-metamorphosed rock samples. Partial Least Squares analysis is used to generate a calibration model from which unknown samples can be analyzed. Principal Components Analysis and Soft Independent Modeling of Class Analogy are employed to generate a model and predict the rock type of the samples. These Multivariate Analysis techniques appear to exploit the matrix effects associated with the chemistries of these 18 samples.     

Chemometric determination of arsenic and lead in untreated powdered red paprika by diffuse reflectance near-infrared spectroscopy

It has been evaluated the potential of near-infrared (NIR) diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) as a way for non-destructive measurement of trace elements at μg kg⁻¹ level in foods, with neither physical nor chemical pre-treatment. Predictive models were developed using partial least-square (PLS) multivariate approaches based on first-order derivative spectra. A critical comparison of two spectral pre-treatments, multiplicative signal correction (MSC) and standard normal variate (SNV) was also made. The PLS models built after using SNV provided the best prediction results for the determination of arsenic and lead in powdered red paprika samples. Relative root-mean-square error of prediction (RRMSEP) of 23% for both metals, arsenic and lead, were found in this study using 20 well characterized samples for calibration and 13 additional samples as validation set. Results derived from this study showed that NIR diffuse reflectance spectroscopy combined with the appropriate chemometric tools could be considered as an useful screening tool for a rapid determination of As and Pb at concentration level of the order of hundred μg kg⁻¹.     

The effect of poly(ethylene glycol) on the activity and structure of glucose-6-phosphate dehydrogenase in solution

The effect of poly(ethylene glycol), PEG, on the enzymatic activity of glucose-6-phosphate dehydrogenase (G-6-PDH) in the oxidation of glucose-6-phosphate (G-6-P), using NADP+ as co-enzyme was investigated. The enzymatic activity was determined by means of spectrophotometry in three different media: pure Tris–HCl buffer, solution of PEG400 (20 wt.%) and of PEG4000 (20 wt.%), both in buffer. Comparing the enzymatic activity values measured in pure buffer with those in the polymer solutions, an increase in the enzymatic activity of 20% was observed in the presence of PEG400 as well as in PEG4000. Calorimetric studies indicated the absence of preferential interactions between G-6-PDH and PEG400 or PEG4000. Nevertheless, the interaction enthalpy, ΔH_int, between NADP+ and PEG400 and PEG4000 amounted to −9.3 and −26.7 kJ/mol, respectively. Small angle X-ray scattering (SAXS) measurements were performed in a higher concentration range. Data analysis performed from SAXS curves by means of the intra-particle distance distribution function p(r) and Guinier plots yielded for G-6-PDH in pure buffer and PEG400 solutions radius of gyration, R_g, of about 70 Å and in PEG4000 solutions, R_g of about 40 Å. The latter has the same dimension as that found in the dimeric crystallographic structure of G-6-PDH, evidencing that G-6-PDH preserves its dimeric configuration in PEG4000 solution. On the contrary, different aggregates of G-6-PDH are formed in the presence of either buffer or PEG400. These findings show that the presence of PEG in solution can exert an effect on the enzyme structure and activity.     

Improving the robustness of a partial least squares (PLS) model based on pure component selectivity analysis and range optimization: case study for the analysis of an etching solution containing hydrogen peroxide

Pure component selectivity analysis (PCSA) was successfully utilized to enhance the robustness of a partial least squares (PLS) model by examining the selectivity of a given component to other components. The samples used in this study were composed of NH₄OH, H₂O₂ and H₂O, a popular etchant solution in the electronic industry. Corresponding near-infrared (NIR) spectra (9000-7500 cm⁻¹) were used to build PLS models. The selective determination of H₂O₂ without influences from NH₄OH and H₂O was a key issue since its molecular structure is similar to that of H₂O and NH₄OH also has a hydroxyl functional group. The best spectral ranges for the determination of NH₄OH and H₂O₂ were found with the use of moving window PLS (MW-PLS) and corresponding selectivity was examined by pure component selectivity analysis. The PLS calibration for NH₄OH was free from interferences from the other components due to the presence of its unique NH absorption bands. Since the spectral variation from H₂O₂ was broadly overlapping and much less distinct than that from NH₄OH, the selectivity and prediction performance for the H₂O₂ calibration were sensitively varied depending on the spectral ranges and number of factors used. PCSA, based on the comparison between regression vectors from PLS and the net analyte signal (NAS), was an effective method to prevent over-fitting of the H₂O₂ calibration. A robust H₂O₂ calibration model with minimal interferences from other components was developed. PCSA should be included as a standard method in PLS calibrations where prediction error only is the usual measure of performance.     

A spectral transfer procedure for application of a single class-model to spectra recorded by different near-infrared spectrometers for authentication of olives in brine

Analytical methods for confirmation of food authenticity claims should be rapid, economic, non-destructive and should not require highly skilled personnel for their deployment. All such conditions are satisfied by spectroscopic techniques. In order to be extensively implemented in routine controls, an ideal method should also give a response independent of the particular equipment used. In the present study, near-infrared (NIR) spectroscopy was used for verifying authenticity of commercial olives in brine of cultivar Taggiasca. Samples were analysed in two laboratories with different NIR spectrometers and a mathematical spectral transfer correction – the boxcar signal transfer (BST) – was developed, allowing to minimise the systematic differences existing between signals recorded with the two instruments. Class models for the verification of olive authenticity were built by the unequal dispersed classes (UNEQ) method, after data compression by disjoint principal component analysis (PCA). Models were validated on an external test set.     

Different strategies for the direct determination of amoxicillin in human urine by second-order multivariate analysis of kinetic-spectrophotometric data

The kinetic evolution of UV-visible absorption spectra of amoxicillin in the presence of copper(II) ions has been processed by the second-order multivariate methods parallel factor analysis (PARAFAC) and also by a novel approach based on partial least-squares with residual bilinearization (PLS/RBL). The latter one is employed for the first time to evaluate kinetic-spectral information. The mechanism of the analyte metal-catalyzed hydrolysis involves a reaction intermediate and a final reaction product, both with spectra which may allow for the determination of amoxicillin in human urine, even in the presence of unsuspected sample components. This is possible thanks to the second-order advantage exploited by the employed chemometric algorithms, among which PARAFAC and PLS/RBL gave the best results. Amoxicillin was determined in a series of spiked and real urine samples, which allowed to perform, respectively, a recovery study and a comparison with the reference high-performance liquid chromatographic technique. The best figures of merit were obtained with PLS/RBL, namely sensitivity, 0.5 AU L mg⁻¹ (AU = absorbance units), analytical sensitivity, 500 L mg⁻¹ and limit of detection, 6 mg L⁻¹. Relative advantages and disadvantages of the employed algorithms are discussed.     

Orthogonal signal correction, wavelet analysis, and multivariate calibration of complicated process fluorescence data

In this paper, multivariate calibration of complicated process fluorescence data is presented. Two data sets related to the production of white sugar are investigated. The first data set comprises 106 observations and 571 spectral variables, and the second data set 268 observations and 3997 spectral variables. In both applications, a single response, ash content, is modelled and predicted as a function of the spectral variables. Both data sets contain certain features making multivariate calibration efforts non-trivial. The objective is to show how principal component analysis (PCA) and partial least squares (PLS) regression can be used to overview the data sets and to establish predictively sound regression models. It is shown how a recently developed technique for signal filtering, orthogonal signal correction (OSC), can be applied in multivariate calibration to enhance predictive power. In addition, signal compression is tested on the larger data set using wavelet analysis. It is demonstrated that a compression down to 4% of the original matrix size — in the variable direction — is possible without loss of predictive power. It is concluded that the combination of OSC for pre-processing and wavelet analysis for compression of spectral data is promising for future use.     

An approach by using near-infrared diffuse reflectance spectroscopy and resin adsorption for the determination of copper, cobalt and nickel ions in dilute solution

Near-infrared spectroscopy (NIRS) has been proved to be a powerful analytical tool and used in various fields, it is seldom, however, used in the analysis of metal ions in solutions. A method for quantitative determination of metal ions in solution is developed by using resin adsorption and near-infrared diffuse reflectance spectroscopy (NIRDRS). The method makes use of the resin adsorption for gathering the analytes from a dilute solution, and then NIRDRS of the adsorbate is measured. Because both the information of the metal ions and their interaction with the functional group of resin can be reflected in the spectrum, quantitative determination is achieved by using multivariate calibration technique. Taking copper (Cu²⁺), cobalt (Co²⁺) and nickel (Ni²⁺) as the analyzing targets and D401 resin as the adsorbent, partial least squares (PLS) model is built from the NIRDRS of the adsorbates. The results show that the concentrations that can be quantitatively detected are as low as 1.00, 1.98 and 1.00 mg L⁻¹ for Cu²⁺, Co²⁺ and Ni²⁺, respectively, and the coexistent ions do not influence the determination.     

Assessing and Advancing Technology for the Noninvasive Measurement of Clinical Glucose

Abstract

Noninvasive glucose sensing is an active area of research and development with many different approaches being explored since the 1980s. The promise of this approach is to provide a measure of the blood glucose concentration in a rapid, painless, and cost-effective manner so that people with diabetes can better maintain tight glycemic control, thereby reducing the complications of diabetes. This mini-review covers the major approaches to noninvasive glucose sensing and focuses on issues that must be resolved before a clinically useful device can be developed. A strong emphasis is placed on measurement selectivity, and methods to examine and characterize the selectivity of such a complex analytical problem are discussed.     

Near-infrared spectroscopy and multivariate calibration for the quantitative determination of certain properties in the petrochemical industry

Near-infrared (NIR) spectroscopy in conjunction with chemometric techniques allows on-line monitoring in real time, which can be of considerable use in industry. If it is to be correctly used in industrial applications, generally some basic considerations need to be taken into account, although this does not always apply. This study discusses some of the considerations that would help evaluate the possibility of applying multivariate calibration in combination with NIR to properties of industrial interest. Examples of these considerations are whether there is a relation between the NIR spectrum and the property of interest, what the calibration constraints are and how a sample-specific error of prediction can be quantified. Various strategies for maintaining a multivariate model after it has been installed are also presented and discussed.     

A feasibility study on the use of visible and short wavelengths in the near-infrared region for the non-destructive measurement of wine composition

The aim of this study was to explore the capability of spectroscopy in the visible (Vis) and short wavelength near-infrared (NIR) regions for the non-destructive measurement of wine composition in intact bottles. In this study we analysed a wide range of commercial wines obtained in Australia in different types of bottles (e.g. colours, diameters and heights), including different wine styles and varieties. Wine bottles were scanned in the Vis-NIR region (600–1,100 nm) in a monochromator instrument in transflectance mode. Principal component analysis (PCA) and partial least-squares (PLS) regression were used to interpret the spectra and develop calibrations for wine composition. Due to the relatively small number of samples available full cross-validation (leave-one-out) was used as validation. The coefficient of correlation in calibration and the standard error of cross-validation (SECV) were 0.67 (SECV: 0.48%), 0.83 (SECV: 4.01 mg L⁻¹), 0.70 (SECV: 28.6 mg L⁻¹) and 0.50 (SECV: 0.15) for alcohol content, total SO₂, free SO₂ and pH, respectively, in the set of wine samples analysed. These preliminary results showed that the assessment of wine composition by Vis and short wavelengths in the NIR is possible for either qualitative analysis (e.g. low-, medium- and high-quality grading), or for screening of composition during bottling and storage. Although low accuracy and precision were obtained for the chemical parameters routinely analysed in wine, calibration models for the chemical parameters were considered acceptable for screening purposes in terms of the standard errors obtained.     

Quantitative analysis of the effect of zidovudine,efavirenz, and ritonavir on insulin aggregation by multivariate curve resolution alternating least squares of infrared spectra

Quantification of the effect of antiretroviral drugs on the insulin aggregation process is an important area of research due to the serious metabolic diseases observed in AIDS patients after prolonged treatment with these drugs. In this work, multivariate curve resolution alternating least squares (MCR-ALS) was applied to infrared monitoring of the insulin aggregation process in the presence of three antiretroviral drugs to quantify their effect. To evidence concentration dependence in this process, mixtures at two different insulin:drug molar ratios were used. The interaction between insulin and each drug was analysed by ¹H NMR spectroscopy. In all cases, the aggregation process was monitored during 45 min by infrared spectroscopy. The aggregates were further characterised by scanning electron microscopy (SEM). MCR-ALS provided the spectral and concentration profiles of the different insulin–drug conformations that are involved in the process. Their feasible band boundaries were calculated using the MCR-BANDS methodology. The kinetic profiles describe the aggregation pathway and the spectral profiles characterise the conformations involved. The retrieved results show that each of the three drugs modifies insulin conformation in a different way, promoting the formation of aggregates. Ritonavir shows the strongest promotion of aggregation, followed by efavirenz and zidovudine. In the studied concentration range, concentration dependence was only observed for zidovudine, with shorter aggregation time obtained as the amount of zidovudine increased. This factor also affected the aggregation pathway.     

Bridging the gap between metabolic profile determination and visualization in neurometabolic disorders: a multivariate analysis of proton magnetic resonance in vivo spectra

The purpose of this work was to check the degree of overlap between rare inborn errors of metabolism and other neurological disorders using principal component analysis of proton magnetic resonance spectroscopy (¹H MRS) in vivo data. We examined 60 patients (median age of 22 months). Fourteen of them were diagnosed with neurometabolic disorders (three cases of metachromatic leukodystrophy, two cases of Canavan disease, two cases of megalencephalic leukoencephalopathy with subcortical cysts, three cases of mitochondrial cytopathy, one case of nonketotic hyperglycinemia, one case of globoid leukodystrophy, one case of congenital disorders of glycosylation, and one case of ethylmalonic encephalopathy). The remaining 46 patients were diagnosed with epilepsy, cerebral palsy, and developmental delay. Results obtained from principal component analysis of complete unresolved ¹H MRS in vivo spectra were interpreted parallelly with LCModel‐derived metabolite levels. The main attention was paid to the following metabolites: N‐acetylaspartate, glutamate + glutamine, creatine, choline, myo‐inositol signal with an uncertain contribution of glycine, and glucose. ¹H MRS in vivo coupled with multivariate analysis is an efficient tool in visualization of metabolic abnormalities in several inborn errors of metabolism (metachromatic leukodystrophy, globoid leukodystrophy, megalencephalic leukoencephalopathy with subcortical cysts, and Canavan disease). Copyright © 2013 John Wiley & Sons, Ltd.     

Complexation of cadmium by the C-terminal hexapeptide Lys-Cys-Thr-Cys-Cys-Ala from mouse metallothionein: study by differential pulse polarography and circular dichroism spectroscopy with multivariate curve resolution analysis

The cadmium-binding properties of the C-terminal hexapeptide of mouse metallothionein I, Lys-Cys-Thr-Cys-Cys-Ala, were studied by circular dichroism spectroscopy (CD), differential pulse polarography (DPP) and ¹¹³Cd-nuclear magnetic resonance (NMR).

The structure of the multiple cadmium binding sites could not be determined by ¹¹³Cd-NMR because of the insolubility of the Cd–peptide samples at the high concentrations required for NMR. Therefore, alternative approaches were used: CD and DPP. The data were analyzed using a multivariate curve resolution (MCR) approach, based on factor analysis techniques, which allows the identification of the signal corresponding to different metal ions bound in different chemical environments. The CD study confirmed that the binding of Cd²⁺ induces important conformational changes in the structure of the peptidic complex, including the formation of a binuclear cluster. The DPP results obtained at various Cd²⁺-to-peptide concentration ratios and pH values, under conditions where electrode adsorption is low, if not negligible, indicated the formation of different Cd²⁺–peptide complexes, and a scheme for the electrochemical reduction of the complexed Cd²⁺ ions is proposed.

These results show that the application of MCR to complex data, such as those from DPP, allows to reach valuable information which is not possible to be obtained by univariate approaches.