Comparison of multivariate curve resolution strategies in quantitative LCxLC: Application to the quantification of furanocoumarins in apiaceous vegetables

Abstract

Comprehensive two-dimensional liquid chromatography (LC × LC) has been gaining popularity for the analysis of complex samples in a wide range of fields including metabolomics, environmental analysis, and food analysis. While LC × LC can provide greater chromatographic resolution than one-dimensional LC (1D-LC), overlapping peaks are often still present in separations of complex samples, a problem that can be alleviated by chemometric curve resolution techniques such as multivariate curve resolution-alternating least squares (MCR-ALS). MCR-ALS has also been previously shown to assist in the quantitative analysis of LC x LC data by isolating pure analyte signals from background signals which are often present at higher levels in LC x LC compared to 1D-LC. In this work we present the analysis of a dataset from the LC × LC analyses of parsley, parsnip and celery samples for the presence and concentrations of 14 furanocoumarins. Several MCR-ALS implementations are compared for the analysis of LC × LC data. These implementations include analyzing the LC x LC chromatogram alone, analyzing the one-dimensional chromatogram alone, as well as two hybrid approaches that make use of both the first and second dimension chromatograms. Furthermore, we compared manual integration of resolved chromatograms versus a simple summation approach, using the resolved chromatographic peaks in both cases. It is found that manual integration of the resolved LC × LC chromatograms provides the best quantification as measured by the consistency between replicate injections. If the summation approach is desired for automation, the choice of MCR-ALS implementation has a large effect on the precision of the analysis. Based on these results, the concentrations of the 14 furanocoumarins are determined in the three apiaceous vegetable types by analyzing the LC × LC chromatograms with MCR-ALS and manual integration for peak area determination. The concentrations of the analytes are found to vary greatly between samples, even within a single vegetable type.

Recent applications of chemometrics in one‐ and two‐dimensional chromatography

“…Other approaches, such as deconvolution or decomposition of the data, may also be applied to improve resolution beyond what is possible given the instrumental constraints. Examples include techniques such as Fourier self-deconvolution [86], wavelets [87], and multivariate curve resolution (MCR) [3]. In many cases, these deconvolution strategies can also be applied for background correction, as they allow one to separate peaks, background, and noise vectors.…”

“…Apart from being a useful preprocessing tool, MCR can also be used to obtain information from chromatograms. Cook et al showed that it is quite useful for quantitation [3,141]. MCR resolves the components of a mixture by deconvoluting the data into response profiles and peak areas.…”

“…Statistical moments in chromatography using trapezoidal and Simpson's rules of peak integration Statistical moments 2019 [127] Linearly modified Gaussian model Curve models 2017 [126] Comparison and optimization of different peak integration methods Integration 2014 [129] Asymmetric least squares baseline algorithm through the accuracy of statistical peak moments Statistical moments 2013 [130] Statistical moments in the exponentially modified Effect of background correction on peak detection and quantification in online comprehensive two-dimensional liquid chromatography AWLS 2012 [36] Comparative study between univariate spectrophotometry and multivariate calibration Comparison 2014 [309] Comparison of multivariate curve resolution strategies in quantitative LC × LC MCR 2017 [3] Quantification with trilinear partial least squares for GC × GC Tri-PLS 2004 [75] Chemometric approach to improve accuracy and precision of quantitation in liquid chromatography MCR 2015 [141] An initial estimation method using cosine similarity for MCR: Application to NMR spectra of chemical mixtures MCR 2019 [39] Determination of phenolic compounds by LC-DAD MCR-ALS 2013 [177] Chromatographic background drift correction coupled with parallel factor analysis to resolve coelution problems in 3D chromatographic data: Quantification of eleven antibiotics in tap water samples by high-performance LC-DAD OSP 2013 [41] Unique resolution of hidden minor peaks in multi-detection chromatography by first-order differentiation and orthogonal projections OSP 1993 [40] Resolution and quantification of peaks in liquid chromatography-mass spectrometry using PARAFAC2 PARAFAC2 2012 [140] Targeted and non-targeted sample profiling by GC × GC-qMS Profiling 2010 [78] Peak detection and quantification of mass spectrometry data acquired from surface-enhanced laser desorption and ionization by denoising spectra with the undecimated discrete wavelet transform…”

“…This multi-channel data may facilitate identification or quantification of the analyte represented by the chromatographic signal. Although co-elution of multiple analytes upon chromatographic analysis may significantly complicate quantification and identification [3,4].…”

Chemometric Strategies for Peak Detection and Profiling from Multidimensional Chromatography

“…Recently, a paper by Cook et al . discussed different strategies based on MCR‐ALS to carry out the quantification of several analytes (furanocoumarins in apiaceous vegetables) in multidimensional data.…”