Search

Introduction: Olive oil, derived from the olive tree (Olea europaea L.), is used in cooking, cosmetics, and soap production. Due to its high value, some producers adulterate olive oil with cheaper edible oils or fraudulently mislabel oils as olive to increase profitability. Adulterated products can cause allergic reactions in sensitive individuals and can lack compounds which contribute to the perceived health benefits of olive oil, and its corresponding premium price. Objective: There is a need for robust methods to rapidly authenticate olive oils. By utilising machine learning models trained on the nuclear magnetic resonance (NMR) spectra of known olive oil and edible oils, samples can be classified as olive and authenticated. While high‐field NMRs are commonly used for their superior resolution and sensitivity, they are generally prohibitively expensive to purchase and operate for routine screening purposes. Low‐field benchtop NMR presents an affordable alternative. Methods: We compared the predictive performance of partial least squares discrimination analysis (PLS‐DA) models trained on low‐field 60 MHz benchtop proton (1H) NMR and high‐field 400 MHz 1H NMR spectra. The data were acquired from a sample set consisting of 49 extra virgin olive oils (EVOOs) and 45 other edible oils. Results: We demonstrate that PLS‐DA models trained on low‐field NMR spectra are highly predictive when classifying EVOOs from other oils and perform comparably to those trained on high‐field spectra. We demonstrated that variance was primarily driven by regions of the spectra arising from olefinic protons and ester protons from unsaturated fatty acids in models derived from data at both field strengths. This study compares the performance of partial least squares discrimination analysis (PLS‐DA) models trained on low‐field (LF) and high‐field (HF) nuclear magnetic resonance (NMR) spectra when distinguishing extra virgin olive oil (EVOO) from other edible oils. When predicting on an external dataset, LF NMR models were highly effective in classifying EVOO and had predictive performance comparable to HF models. These findings support LF NMR trained machine learning algorithms as a rapid, cost‐effective tool for EVOO authentication. [ABSTRACT FROM AUTHOR]

Discriminant-type analyses arise from the need to classify samples based on their measured characteristics (variables), usually with respect to some observable property. In the case of samples that are difficult to obtain, or using advanced instrumentation, it is very common to encounter situations with many more measured characteristics than samples. The method of Partial Least Squares Regression (PLS-R), and its variant for discriminant-type analyses (PLSDA) are among the most ubiquitous of these tools. PLS utilises a rank-deficient method to solve the inverse least-squares problem in a way that maximises the co-variance between the known properties of the samples (commonly referred to as the Y-Block), and their measured characteristics (the X-block). A relatively small subset of highly co-variate variables are weighted more strongly than those that are poorly co-variate, in such a way that an illposed matrix inverse problem is circumvented. Feature selection is another common way of reducing the dimensionality of the data to a relatively small, robust subset of variables for use in subsequent modelling. The utility of these features can be inferred and tested any number of ways, this are the subject of this review. [ABSTRACT FROM AUTHOR]

Comprehensive gas chromatography with time of flight mass spectrometry is a powerful tool in the analysis of complex samples. Chemometric analysis of raw chromatographic data is more useful in one‐ and two‐dimensional separations relative to peak tables. The data volume from such experiments generally necessitates the use of data reduction tools. Such tools often sacrifice some of the multivariate information in the mass to charge ratio dimension. The unique ion filter reduces the over‐redundancy in two‐dimensional gas chromatography‐mass spectrometry data by limiting the data to a few unique/pseudo‐unique ions, sub‐peaks/slices in the first dimension, and spectra in the second dimension. We explore the performance of this algorithm through careful inspection of two‐dimensional gas chromatography‐mass spectrometry data before and after application of the filter. A reduction (99%) in the number of variables in a two‐dimensional gas chromatography‐mass spectrometry chromatogram passed on to subsequent analysis was observed. Feature selection times for model optimization reduced from 229 (±13) to 6.8 (±0.5) min when the filter was applied. An estimate of two unique/pseudo‐unique ions, one sub‐peak in the first dimension and five spectra in the second dimension were considered to provide a true representation of each chromatogram and provided enough information to achieve 100% model prediction accuracy. [ABSTRACT FROM AUTHOR]

Abstract: A three-dimensional extension of a previously developed metric termed cluster resolution is presented. The cluster resolution metric considers confidence ellipses (here three-dimensional confidence ellipsoids) around clusters of points in principal component or latent variable space. Cluster resolution is defined as the maximum confidence limit at which confidence ellipses do not overlap and can serve to guide automated variable selection processes. Previously, this metric has been used to guide variable selection in a two-dimensional projection of data. In this study, the metric is refined to simultaneously consider the shapes of clusters of points in a three-dimensional space. We couple it with selectivity ratio-based variable ranking and a combined backward elimination/forward selection strategy to demonstrate its use for the automated optimization of a six-class PCA model of gasoline by vendor and octane rating. Within-class variability was artificially increased through evaporative weathering and intentional contamination of samples, making the optimization more challenging. Our approach was successful in identifying a small subset of variables (644) from the raw GC–MS chromatographic data which comprised ∼2×106 variables per sample. In the final model there was clear separation between all classes. Computational time for this completely automated variable selection was 36h; slower than solving the same problem using three two-dimensional projections, but yielding an overall better model. By simultaneously considering three dimensions instead of only two at a time, the resulting overall cluster resolution was improved. [Copyright &y& Elsevier]

Abstract: A novel metric termed cluster resolution is presented. This metric compares the separation of clusters of data points while simultaneously considering the shapes of the clusters and their relative orientations. Using cluster resolution in conjunction with an objective variable ranking metric allows for fully automated feature selection for the construction of chemometric models. The metric is based upon considering the maximum size of confidence ellipses around clusters of points representing different classes of objects that can be constructed without any overlap of the ellipses. For demonstration purposes we utilized PCA to classify samples of gasoline based upon their octane rating. The entire GC–MS chromatogram of each sample comprising over 2×106 variables was considered. As an example, automated ranking by ANOVA was applied followed by a forward selection approach to choose variables for inclusion. This approach can be generally applied to feature selection for a variety of applications and represents a significant step towards the development of fully automated, objective construction of chemometric models. [ABSTRACT FROM AUTHOR]

We have previously demonstrated the successful application of multivariate statistical (i.e. chemometric) techniques to the classification of casework fire debris based on gasoline content, with the objective to expedite the data interpretation process for the forensic analysis of fire debris. We have also shown that it is possible to classify simulated fire debris based on gasoline content using simulated fire debris; however, models trained on simulated debris are not applicable to casework samples without a significant loss in model accuracy. A previously developed simulation protocol that works well for generating debris to train human analysts was inadequate for training either partial least-squares discriminant analysis (PLS-DA) or soft independent modelling by class analogy (SIMCA) models to identify casework debris due mainly to its inability to generate a sufficient amount of benzene, toluene, ethylbenzene, and xylenes (BTEX) and non-aromatic hydrocarbon compounds. This method relied on pyrolyzing materials at 400 °C. Here we examine the effects of pyrolysis conditions on household materials, including spruce plywood, vinyl sheet flooring, polyethylene terephthalate (PET) carpet, Nylon 6 carpet, polyurethane (PU) foam carpet underlay, asphalt shingle, medium-density fireboard (MDF) shelving, and spruce timber at temperatures above 400 °C, in an attempt to generate additional BTEX and non-aromatic hydrocarbon compounds for the realistic simulation of fire debris. The work presented here showed that C 3 - to C 5 -alkylbenzenes, which are abundant in gasoline, were generally absent from the pyrolysates of all materials studied and at all temperatures studied (400, 700 and 900 °C), only appearing in trace amounts on rare occasions. Based on our results, we propose that accurate simulation of fire debris could be achieved with a mixture of carpets, carpet underlay, and vinyl flooring pyrolyzed at 700 °C; spruce plywood pyrolyzed at 900 °C, and asphalt shingles pyrolyzed at 400 and 700 °C. These conditions generated a substantial amount of BTEX and non-aromatic hydrocarbons in the debris matrix background. The testing of this “recipe” for generating simulated debris for training chemometric models that can classify casework debris is left for future experimentation. [ABSTRACT FROM AUTHOR]

Detection and identification of ignitable liquids (ILs) in arson debris is a critical part of arson investigations. The challenge of this task is due to the complex and unpredictable chemical nature of arson debris, which also contains pyrolysis products from the fire. ILs, most commonly gasoline, are complex chemical mixtures containing hundreds of compounds that will be consumed or otherwise weathered by the fire to varying extents depending on factors such as temperature, air flow, the surface on which IL was placed, etc. While methods such as ASTM E-1618 are effective, data interpretation can be a costly bottleneck in the analytical process for some laboratories. In this study, we address this issue through the application of chemometric tools. Prior to the application of chemometric tools such as PLS-DA and SIMCA, issues of chromatographic alignment and variable selection need to be addressed. Here we use an alignment strategy based on a ladder consisting of perdeuterated n-alkanes. Variable selection and model optimization was automated using a hybrid backward elimination (BE) and forward selection (FS) approach guided by the cluster resolution (CR) metric. In this work, we demonstrate the automated construction, optimization, and application of chemometric tools to casework arson data. The resulting PLS-DA and SIMCA classification models, trained with 165 training set samples, have provided classification of 55 validation set samples based on gasoline content with 100% specificity and sensitivity. [ABSTRACT FROM AUTHOR]

Abstract: Petroleum analysis presents many unique challenges as a result of the overwhelming number of compounds present in petroleum samples. Consequently the use of multidimensional separation techniques will almost invariably be required in order to overcome these challenges. Within this paper we review recent developments in the application of comprehensive multidimensional techniques for petroleum analysis focusing on more recent applications. Basic instrumentation for various comprehensive multidimensional techniques is outlined along with an overview of a broad range of applications in both group-type and target molecule analyses for petroleum and biofuel analysis. In addition, strategies for data interpretation and chemometric analysis of multidimensional data are also reviewed. [Copyright &y& Elsevier]

Olive oil is the oil obtained from the fruit of the olive tree, Olea europaea L., and is commonly used in cooking, cosmetics, and soaps. The high value and demand for olive oil products has led to some producers adding cheaper edible non-olive oils to olive oil to increase their profitability. Many of these adulterant oils have similar taste profiles to extra virgin olive oil; however, these can cause allergic reactions in individuals with sensitivities to the non-olive oil. Given the implications of olive oil adulteration, there is a demand for methods to identify and quantify adulterant oils in olive oil samples. This work explores the utility of low-field benchtop nuclear magnetic resonance (NMR) with a novel chemometrics workflow for the quantification of non-olive oils in authentic extra virgin olive oil samples. The measurements were obtained using a 60 MHz 1H NMR. The icoshift algorithm was utilized for alignment, and partial least squares regression models of the entire NMR spectrum were used for quantification. Here, 17 different varieties of edible non-olive oils were investigated. The models were highly linear, accurate, and precise. Most of these adulterant oils were detectable below 10% v/v in extra virgin olive oil samples. • Benchtop 60 MHz NMR for quantification of adulterant oils in olive oil. • Use of chemometrics to align and generate partial least squares regression models. • 17 different non-olive oil vs. olive oil models generated. • Detection of adulterants below 10% v/v for most adulterant oils. [ABSTRACT FROM AUTHOR]

• Ratio between first and second Eigenvalues from GC–MS data serves as an approximation of an F-value, allowing them to be interpreted as Fisher ratios. • A moving window across a GC–MS chromatogram is used to compute Fisher ratios. • Probability of each acquisition being in a region of interest is computed from Fisher ratios. • Our algorithm is capable of detecting chromatographic regions of interest down to 5 pg on column. • Free, open-source algorithm available for implementation into existing chemometric workflows. There are many challenges associated with analysing gas chromatography - mass spectrometry (GC–MS) data. Many of these challenges stem from the fact that electron ionization (EI) can make it difficult to recover molecular information due to the high degree of fragmentation with concomitant loss of molecular ion signal. With GC–MS data there are often many common fragment ions shared among closely-eluting peaks, necessitating sophisticated methods for analysis. Some of these methods are fully automated, but make some assumptions about the data which can introduce artifacts during the analysis. Chemometric methods such as Multivariate Curve Resolution (MCR), or Parallel Factor Analysis (PARAFAC/PARAFAC2) are particularly attractive, since they are flexible and make relatively few assumptions about the data - ideally resulting in fewer artifacts. These methods do require expert user intervention to determine the most relevant regions of interest and an appropriate number of components, k , for each region. Automated region of interest selection is needed to permit automated batch processing of chromatographic data with advanced signal deconvolution. Here, we propose a new method for automated, untargeted region of interest selection that accounts for the multivariate information present in GC–MS data to select regions of interest based on the ratio of the squared first, and second singular values from the Singular Value Decomposition (SVD) of a window that moves across the chromatogram. Assuming that the first singular value accounts largely for signal, and that the second singular value accounts largely for noise, it is possible to interpret the relationship between these two values as a probabilistic distribution of Fisher Ratios. The sensitivity of the algorithm was tested by investigating the concentration at which the algorithm can no longer pick out chromatographic regions known to contain signal. The algorithm achieved detection of features in a GC–MS chromatogram at concentrations below 10 pg on-column. The resultant probabilities can be interpreted as regions that contain features of interest. [ABSTRACT FROM AUTHOR]

Abstract: Direct chemometric interpretation of raw chromatographic data (as opposed to integrated peak tables) has been shown to be advantageous in many circumstances. However, this approach presents two significant challenges: data alignment and feature selection. In order to interpret the data, the time axes must be precisely aligned so that the signal from each analyte is recorded at the same coordinates in the data matrix for each and every analyzed sample. Several alignment approaches exist in the literature and they work well when the samples being aligned are reasonably similar. In cases where the background matrix for a series of samples to be modeled is highly variable, the performance of these approaches suffers. Considering the challenge of feature selection, when the raw data are used each signal at each time is viewed as an individual, independent variable; with the data rates of modern chromatographic systems, this generates hundreds of thousands of candidate variables, or tens of millions of candidate variables if multivariate detectors such as mass spectrometers are utilized. Consequently, an automated approach to identify and select appropriate variables for inclusion in a model is desirable. In this research we present an alignment approach that relies on a series of deuterated alkanes which act as retention anchors for an alignment signal, and couple this with an automated feature selection routine based on our novel cluster resolution metric for the construction of a chemometric model. The model system that we use to demonstrate these approaches is a series of simulated arson debris samples analyzed by passive headspace extraction, GC–MS, and interpreted using partial least squares discriminant analysis (PLS-DA). [Copyright &y& Elsevier]