Resolution of Complex Liquid Chromatography−Fourier Transform Infrared Spectroscopy Data

A Novel Variable Selection Method Based on Binning-Normalized Mutual Information for Multivariate Calibration

Variable (wavelength) selection is essential in the multivariate analysis of near-infrared spectra to improve model performance and provide a more straightforward interpretation. This paper proposed a new variable selection method named binning-normalized mutual information (B-NMI) based on information entropy theory. "Data binning" was applied to reduce the effects of minor measurement errors and increase the features of near-infrared spectra. "Normalized mutual information" was employed to calculate the correlation between each wavelength and the reference values. The performance of B-NMI was evaluated by two experimental datasets (ideal ternary solvent mixture dataset, fluidized bed granulation dataset) and two public datasets (gasoline octane dataset, corn protein dataset). Compared with classic methods of backward and interval PLS (BIPLS), variable importance projection (VIP), correlation coefficient (CC), uninformative variables elimination (UVE), and competitive adaptive reweighted sampling (CARS), B-NMI not only selected the most featured wavelengths from the spectra of complex real-world samples but also improved the stability and robustness of variable selection results.

Recent Advances in Standardization of Herbal Drugs

Background:

Herbal drugs play a significant role to maintain the human healthiness and to treat the ailments since the dawn of civilization. Moreover, these plants have provided many lead compounds that culminated in modern medicine. A single herb is regarded as mini-combinatorial library of phytoconstituents hence the quality control of herbal drugs in an herbal formulation is not an easy task because a number of factors impact their pharmacological efficiency and consistent therapeutic effects. Hence, to provide consistent beneficial therapeutic effects, standardized herbal products of consistent quality and purity are required.

Methods:

This review is based on publications obtained by a selective search in PubMed using the keywords “Standardized herbal products”, “fingerprinting”, “authentication”, “chemometric, hyphenated techniques”, “quality control of herbal drugs”, “identification”.

Results:

In the era of modernization, chromatographic techniques coupled with sophisticated spectroscopic analytical methods are used in estimating the authenticity, identity and characteristic of herbal products. Further, with the advancement of computer technology, chemometrics methods have become a leading tool with an unsupervised pattern recognition technique for handling multivariate data without prior knowledge about the studied samples and mines more beneficial and valuable information about the chemical entities from the raw data.

Conclusion:

Standardization of HDs chromatographic fingerprint is not always a perfect way to present all compounds. To assess the quality of medicinal plants, new ways are regularly being explored such as combination chemical fingerprint with biological methods, biofingerprint and metabolic fingerprint quality metrology, pharmacodynamics and export system of medicinal plants have been researched in some groups but still a significant amount of work is required to achieve a perfect system for quality evaluation of herbal drugs. Further, novel chemometric techniques have been unfolded that mines more beneficial and valuable information about the chemical entities from the raw data. So this review emphasis mainly on hyphenated techniques associated with chemometric method used in herbal drugs for identifying more valuable information and various methods for providing data, among which most commonly used techniques are chemometric resolution method and Principal Component Analysis (PCA) method.

Chemometrics: A new scenario in herbal drug standardization

Chromatography and spectroscopy techniques are the most commonly used methods in standardization of herbal medicines but the herbal system is not easy to analyze because of their complexity of chemical composition. Many cutting-edge analytical technologies have been introduced to evaluate the quality of medicinal plants and significant amount of measurement data has been produced. Chemometric techniques provide a good opportunity for mining more useful chemical information from the original data. Then, the application of chemometrics in the field of medicinal plants is spontaneous and necessary. Comprehensive methods and hyphenated techniques associated with chemometrics used for extracting useful information and supplying various methods of data processing are now more and more widely used in medicinal plants, among which chemometrics resolution methods and principal component analysis (PCA) are most commonly used techniques. This review focuses on the recent various important analytical techniques, important chemometrics tools and interpretation of results by PCA, and applications of chemometrics in quality evaluation of medicinal plants in the authenticity, efficacy and consistency.

Improved modeling of in vivo confocal Raman data using multivariate curve resolution (MCR) augmentation of ordinary least squares models

In vivo confocal Raman spectroscopy has become the measurement technique of choice for skin health and skin care related communities as a way of measuring functional chemistry aspects of skin that are key indicators for care and treatment of various skin conditions. Chief among these techniques are stratum corneum water content, a critical health indicator for severe skin condition related to dryness, and natural moisturizing factor components that are associated with skin protection and barrier health. In addition, in vivo Raman spectroscopy has proven to be a rapid and effective method for quantifying component penetration in skin for topically applied skin care formulations. The benefit of such a capability is that noninvasive analytical chemistry can be performed in vivo in a clinical setting, significantly simplifying studies aimed at evaluating product performance. This presumes, however, that the data and analysis methods used are compatible and appropriate for the intended purpose. The standard analysis method used by most researchers for in vivo Raman data is ordinary least squares (OLS) regression. The focus of work described in this paper is the applicability of OLS for in vivo Raman analysis with particular attention given to use for non-ideal data that often violate the inherent limitations and deficiencies associated with proper application of OLS. We then describe a newly developed in vivo Raman spectroscopic analysis methodology called multivariate curve resolution-augmented ordinary least squares (MCR-OLS), a relatively simple route to addressing many of the issues with OLS. The method is compared with the standard OLS method using the same in vivo Raman data set and using both qualitative and quantitative comparisons based on model fit error, adherence to known data constraints, and performance against calibration samples. A clear improvement is shown in each comparison for MCR-OLS over standard OLS, thus supporting the premise that the MCR-OLS method is better suited for general-purpose multicomponent analysis of in vivo Raman spectral data. This suggests that the methodology is more readily adaptable to a wide range of component systems and is thus more generally applicable than standard OLS.

Problem of mixtures with known compositions and IRONFLEA method for multivariate curve resolution

A special case of gray spectral data systems [(a) F.-T. Chau, Y.-Z. Liang, J. Gao, X.-G. Shao (Eds.), Chemometrics: From Basics to Wavelet Transform, Chemical Analysis Series, vol. 164, John Wiley & Sons, Inc., 2004; (b) Y.Z. Liang, O.M. Kvalheim, R. Manne, Chemom. Intell. Lab. Syst. 18 (1993) 235-250] is discussed here and the least-squares method for the multivariate curve resolution (MCR) named IRONFLEA is proposed. The system under consideration is the bilinear spectral data of the samples with known chemical compositions and unknown concentration matrix. If the spectra of samples (A(i)) and (Q+A(i)) (i = 1, ..., n, n > or = 2) are available, then the spectrum and the concentrations of Q could be found and the solution is unique. A practical chemical model for this problem could be mixtures, polymers, peptides, oligosaccharides, or supramolecular formations made of a limited number of monomeric components. In the cases of polymeric or oligomeric samples the spectral contributions and the concentrations of the particular monomeric units are extracted. The method is capable of extracting chemically meaningful spectra of components. The method is implemented in SAS IML code and tested for the deconvolution of spectra of polymers made of styrene derivatives with known monomeric compositions [(a) H. Fenniri, L. Ding, A.E. Ribbe, Y. Zyrianov, J. Am. Chem. Soc. 123 (2001) 8151-8152; (b) H. Fenniri, S. Chun, L. Ding, Y. Zyrianov, K. Hallenga, J. Am. Chem. Soc. 125 (2003) 10546-10560]. The method performs calculations fast enough to allow the incorporation of leave-one-out outlier removal procedure.

Factor analysis of hyphenated chromatographic data

Factor analysis (FA) is a family of widely used methods to obtain the underlying sources of variation of data tables. Typically, hyphenated chromatographic data provide data tables with one elution direction and another linked to the detector response. In this context, the factors are the eluting compounds and the profiles defining each factor are the elution profile and the pure response of the compound. This article describes the use of FA in chromatography through diverse tools and problems. Examples of determination of number of compounds, peak purity problems, resolution of overlapped compounds or extension to simultaneous analysis of multiple runs (higher-order data structures) to obtain qualitative and quantitative information are reviewed.

Towards unsupervised analysis of second-order chromatographic data: automated selection of number of components in multivariate curve-resolution methods

A method to apply multivariate curve-resolution unattendedly is presented. The algorithm is suitable to perform deconvolution of two-way data (e.g. retrieving the individual elution profiles and spectra of co-eluting compounds from signals obtained from a chromatograph equipped with multiple-channel detection: LC-DAD or GC-MS). The method is especially adequate to achieve the advantages of deconvolution approaches when huge amounts of data are present and manual application of multivariate techniques is too time-consuming. The philosophy of the algorithm is to mimic the reactions of an expert user when applying the orthogonal projection approach--multivariate curve-resolution techniques. Basically, the method establishes a way to check the number of significant components in the data matrix. The performance of the method was superior to the Malinowski F-test. The algorithm was tested with HPLC-DAD signals.

On-line HPLC combined with multivariate statistical process control for the monitoring of reactions

On-line high performance liquid chromatography is used to monitor a steady state reaction over 35.2 h, with 197 chromatograms recorded as the reaction progresses. For each chromatogram, peaks are detected, baseline corrected, aligned and integrated to provide a peak table consisting of the intensities of 19 peaks, two corresponding to the reactants, one to the product and one to the solvent, the remaining being impurities, by-products or intermediates. D-charts and Q-charts from multivariate statistical process control are applied to the data to determine which samples are out of control and also provide diagnostic insight into why these samples are problematic. The D-chart is best at looking at overall performance issues such as problems with mixing or difficulties with instrument operation, whereas the Q-charts are best at detecting impurities during the reaction.

Towards the solution of the eluent elimination problem in high-performance liquid chromatography–infrared spectroscopy measurements by chemometric methods

The high-performance liquid chromatography-infrared spectroscopy (HPLC-IR) technique utilizing on-line flow through cell (FTC) detection has an inherent practical problem: strong absorption bands of the eluent may mask valuable analytical regions of the IR spectrum. The experimentalists' answer to this challenge is physical elimination of the chromatographic eluent before spectroscopic detection, which however results in off-line measurement of spectra. In the present work, the capabilities of some chemometric algorithms using iteratively applied multi-way methods such as parallel factor analysis (PARAFAC) and PARAFAC2, developed with the aim of overcoming the problems of eluent elimination are examined and evaluated. Test calculations done on simulated liquid chromatographic infrared (LC-IR) data cubes have shown that although PARAFAC2 performs much better than the simple PARAFAC method, it does not give correct decompositions, just like multivariate curve resolution with alternative least squares (MCR-ALS) and related bilinear data based methods. In search for a better solution, a method named objective subtraction of solvent spectrum with iterative use of PARAFAC and PARAFAC2 (OSSS-IU-PARAFAC and OSSS-IU-PARAFAC2) has been developed. Calculations performed with the corresponding Matlab program developed by the authors and run with the appropriate functions in PLS_Toolbox yielded very promising results in evaluations of both simulated and real HPLC-IR data sets, after necessary data pretreatments.

Application of multivariate curve resolution methods to on-flow LC-NMR

The application of evolving window factor analysis (EFA), subwindow factor analysis (SFA), iterative target transformation factor analysis (ITTFA), alternating least squares (ALS), Gentle, automatic window factor analysis (AUTOWFA) and constrained key variable regression (CKVR) to resolve on-flow LC-NMR data of eight compounds into individual concentration and spectral profiles is described. CKVR has been reviewed critically and modifications are suggested to obtain improved results. A comparison is made between three single variable selection methods namely, orthogonal projection approach (OPA), simple-to-use interactive self-modelling mixture analysis approach (SIMPLISMA) and simplified Borgen method (SBM). It is demonstrated that LC-NMR data can be resolved if NMR peak cluster information is utilised.

Application of GRAM and TLD to the resolution and quantitation of real complex multicomponent mixtures by fluorescence spectroscopy

The application of the generalised rank annihilation method (GRAM) and the trilinear decomposition (TLD) method to the resolution and quantitation of fluorescence excitation-emission matrices of a ternary mixture of pesticides, carbendazim, fuberidazole, and thiabendazole, with overlapped spectra is described. The results obtained with both methods are compared and evaluated using measures of similarity (correlation coefficients) between the real and estimated spectra. Both approaches have been tested using augmented data matrices containing only two samples, but none of these methods succeeded completely in resolution of the system. When TLD was applied to augmented data matrices containing more than two samples better performance was achieved. To illustrate the application of both algorithms to real samples, they were used in the analysis of water samples containing the target pesticides. Again, TLD had an advantage over GRAM because the ability to analyse data from multiple (more than two) samples simultaneously allowed the resolution of the mixtures.

Application of multivariate curve resolution for analysis of FT-IR microspectroscopic images of in situ plant tissue

The chemometric techniques of multivariate curve resolution (MCR) are aimed at extracting the spectra and concentrations of individual components present in mixtures using a minimum set of initial assumptions. We present results from the application of alternating least squares (ALS) based MCR to the analysis of hyperspectral images of in situ biological material. The spectra of individual pure components were mathematically extracted and then identified by searching the spectra against a commercial library. No prior information about the chemical composition of the material was used in the data analysis. The spectra recovered by ALS-MCR analysis of an FT-IR microspectroscopic image of an 8-micron-cornkernel section matched very well the spectra of the corn storage protein, zein, and starch. Through the application of MCR, we were able to show the presence of a second spectrally different protein, which could not be easily seen using univariate analysis. These results demonstrate the value of multivariate curve resolution techniques for the analysis of biological tissue. The value of principal components analysis (PCA) for hyperspectral image analysis is also discussed.

Use of the orthogonal projection approach (OPA) to monitor batch processes

The orthogonal projection approach (OPA) and multivariate curve resolution (MCR) are presented as a way to monitor batch processes using spectroscopic data. Curve resolution allows one to look within a batch and predict on-line real concentration profiles of the different species appearing during reactions. Taking into account the variations of the process by using an augmented matrix of complete batches, the procedure explained here calculates some prediction coefficients that can afterwards be applied for a new batch.

Determination of ebrotidine metabolites in overlapping peaks from capillary zone electrophoresis using chemometric methods

This paper illustrates the possibilities of chemometric methods in the resolution and quantification of various compounds in overlapping peaks from capillary electrophoresis. Ebrotidine and most of its metabolites were efficiently separated by capillary zone electrophoresis (CZE) in a fused-silica capillary. However, the procedure was not suitable for the physical separation of the three less ionizable metabolites, which comigrated and overlapped with the electroosmotic flow signal. Multivariate curve resolution based on an alternating least squares procedure was used for their mathematical resolution. For such a purpose, data obtained in the CZE system with a diode array detector, which consisted of UV spectra registered over time, were analyzed. The ebrotidine metabolites were successfully resolved and quantified in synthetic mixtures and urine samples.

Determination of amino acids in overlapped capillary electrophoresis peaks by means of partial least-squares regression

Amino acid derivatives of 1,2-naphthoquinone-4-sulfonate (NQS) can be separated by capillary electrophoresis at 30 kV in a fused-silica capillary by using a 40 mM sodium tetraborate-isopropanol (3:1, v/v) solution as background electrolyte. This procedure was suitable for the most common amino acids. However, the peaks of three amino acids (phenylalanine, isoleucine and tyrosine) were only partially resolved and peaks of histidine and leucine derivatives overlapped completely. Partial least-squares regression (PLS) may overcome the lack of selectivity for these amino acids. Spectroelectropherograms of the corresponding amino acid derivative peaks were monitored with a diode-array spectrophotometer in the range 225 to 540 nm. Both spectra and electropherograms can be used as multivariate data for further analysis. In general, the best predictions were obtained using the time domain.

Resolution of overlapped peaks of amino acid derivatives in capillary electrophoresis using multivariate curve resolution based on alternating least squares

The application of chemometric techniques to the resolution of overlapped peaks in capillary electrophoresis (CE) is described. When a physical separation can not be completely accomplished, chemometrics might still resolve the determination of the analytes mathematically. CE with diode array detection can provide a large amount of data consisting of spectra registered over time. In this study, the capillary electrophoretic separation of 1,2-naphthoquinone-4-sulfonate derivatives of amino acids is studied. Most of the common amino acid derivatives can be separated at 30 kV in a fused-silica capillary by using a 40 mM sodium tetraborate + isopropanol (3:1 v/v) solution as background electrolyte. However, peaks of certain derivatives (Phe, His, Leu and Ile) still overlap. A multivariate curve resolution method based on an alternating least squares optimization procedure is used for the resolution of the overlapped electrophoretic peaks. The method takes advantage of spectral and electrophoretic differences of analytes to recover their pure electrophoretic and spectral profiles. In addition, each analyte in the mixture can be quantified using the corresponding standards.