Investigating the use of visible and near infrared spectroscopy to predict sensory and texture attributes of beef M. longissimus thoracis et lumborum.

The aim of this study was to calibrate chemometric models to predict beef M. longissimus thoracis et lumborum (LTL) sensory and textural values using visible-near infrared (VISNIR) spectroscopy. Spectra were collected on the cut surface of LTL steaks both on-line and off-line. Cooked LTL steaks were analysed by a trained beef sensory panel as well as undergoing WBSF analysis. The best coefficients of determination of cross validation (R2CV) in the current study were for textural traits (WBSF = 0.22; stringiness = 0.22; crumbly texture = 0.41: all 3 models calibrated using 48 h post-mortem spectra), and some sensory flavour traits (fatty mouthfeel = 0.23; fatty after-effect = 0.28: both calibrated using 49 h post-mortem spectra). The results of this experiment indicate that VISNIR spectroscopy has potential to predict a range of sensory traits (particularly textural traits) with an acceptable level of accuracy at specific post-mortem times.

Online Prediction of Physico-Chemical Quality Attributes of Beef Using Visible-Near-Infrared Spectroscopy and Chemometrics.

The potential of visible-near-infrared (Vis-NIR) spectroscopy to predict physico-chemical quality traits in 368 samples of bovine musculus longissimus thoracis et lumborum (LTL) was evaluated. A fibre-optic probe was applied on the exposed surface of the bovine carcass for the collection of spectra, including the neck and rump (1 h and 2 h post-mortem and after quartering, i.e., 24 h and 25 h post-mortem) and the boned-out LTL muscle (48 h and 49 h post-mortem). In parallel, reference analysis for physico-chemical parameters of beef quality including ultimate pH, colour (L, a*, b*), cook loss and drip loss was conducted using standard laboratory methods. Partial least-squares (PLS) regression models were used to correlate the spectral information with reference quality parameters of beef muscle. Different mathematical pre-treatments and their combinations were applied to improve the model accuracy, which was evaluated on the basis of the coefficient of determination of calibration (R2C) and cross-validation (R2CV) and root-mean-square error of calibration (RMSEC) and cross-validation (RMSECV). Reliable cross-validation models were achieved for ultimate pH (R2CV: 0.91 (quartering, 24 h) and R2CV: 0.96 (LTL muscle, 48 h)) and drip loss (R2CV: 0.82 (quartering, 24 h) and R2CV: 0.99 (LTL muscle, 48 h)) with lower RMSECV values. The results show the potential of Vis-NIR spectroscopy for online prediction of certain quality parameters of beef over different time periods.

Performances of full cross-validation partial least squares regression models developed using Raman spectral data for the prediction of bull beef sensory attributes.

The data presented in this article are related to the research article entitled "Application of Raman spectroscopy and chemometric techniques to assess sensory characteristics of young dairy bull beef" [1]. Partial least squares regression (PLSR) models were developed on Raman spectral data pre-treated using Savitzky Golay (S.G.) derivation (with 2nd or 5th order polynomial baseline correction) and results of sensory analysis on bull beef samples (n = 72). Models developed using selected Raman shift ranges (i.e. 250-3380 cm-1, 900-1800 cm-1 and 1300-2800 cm-1) were explored. The best model performance for each sensory attributes prediction was obtained using models developed on Raman spectral data of 1300-2800 cm-1.

Assessment of infant formula quality and composition using Vis-NIR, MIR and Raman process analytical technologies.

In this study, visible and near-infrared (Vis-NIR), mid-infrared (MIR) and Raman process analytical technologies were investigated for assessment of infant formula quality and compositional parameters namely preheat temperature, storage temperature, storage time, fluorescence of advanced Maillard products and soluble tryptophan (FAST) index, soluble protein, fat and surface free fat (SFF) content. PLS-DA models developed using spectral data with appropriate data pre-treatment and significant variables selected using Martens' uncertainty test had good accuracy for the discrimination of preheat temperature (92.3-100%) and storage temperature (91.7-100%). The best PLS regression models developed yielded values for the ratio of prediction error to deviation (RPD) of 3.6-6.1, 2.1-2.7, 1.7-2.9, 1.6-2.6 and 2.5-3.0 for storage time, FAST index, soluble protein, fat and SFF content prediction respectively. Vis-NIR, MIR and Raman were demonstrated to be potential PAT tools for process control and quality assurance applications in infant formula and dairy ingredient manufacture.

Application of Raman spectroscopy and chemometric techniques to assess sensory characteristics of young dairy bull beef.

This work aims to develop a rapid analytical technique to predict beef sensory attributes using Raman spectroscopy (RS) and to investigate correlations between sensory attributes using chemometric analysis. Beef samples (n = 72) were obtained from young dairy bulls (Holstein-Friesian and Jersey×Holstein-Friesian) slaughtered at 15 and 19 months old. Trained sensory panel evaluation and Raman spectral data acquisition were both carried out on the same longissimus thoracis muscles after ageing for 21 days. The best prediction results were obtained using a Raman frequency range of 1300-2800 cm-1. Prediction performance of partial least squares regression (PLSR) models developed using all samples were moderate to high for all sensory attributes (R2CV values of 0.50-0.84 and RMSECV values of 1.31-9.07) and were particularly high for desirable flavour attributes (R2CVs of 0.80-0.84, RMSECVs of 4.21-4.65). For PLSR models developed on subsets of beef samples i.e. beef of an identical age or breed type, significant improvements on prediction performances were achieved for overall sensory attributes (R2CVs of 0.63-0.89 and RMSECVs of 0.38-6.88 for each breed type; R2CVs of 0.52-0.89 and RMSECVs of 0.96-6.36 for each age group). Chemometric analysis revealed strong correlations between sensory attributes. Raman spectroscopy combined with chemometric analysis was demonstrated to have high potential as a rapid and non-destructive technique to predict the sensory quality traits of young dairy bull beef.

Assessment of physico-chemical traits related to eating quality of young dairy bull beef at different ageing times using Raman spectroscopy and chemometrics.

Raman spectroscopy and chemometrics were investigated for the prediction of eating quality related physico-chemical traits of Holstein-Friesian bull beef. Raman spectra were collected on the 3rd, 7th and 14th days post-mortem. A frequency range of 1300-2800cm-1 was used for partial least squares (PLS) modelling. PLS regression (PLSR) models for the prediction of WBSF and cook loss achieved an R2CV of 0.75 with RMSECV of 6.82 N and an R2CV of 0.77 with RMSECV of 0.97%w/w respectively. For the prediction of intramuscular fat, moisture and crude protein content, R2CV values were 0.85, 0.91 and 0.70 with RMSECV of 0.52%w/w, 0.39%w/w and 0.38%w/w respectively. An R2CV of 0.79 was achieved for the prediction of both total collagen and hydroxyproline content, while for collagen solubility the R2CV was 0.88. All samples (100%) from 15- and 19-month old bulls were correctly classified using PLS discriminant analysis (PLS-DA), while 86.7% of samples from different muscles (longissimus thoracis, semitendinosus and gluteus medius) were correctly classified. In general, PLSR models using Raman spectra on the 3rd day post-mortem had better prediction performance than those on the 7th and 14th days. Raman spectroscopy and chemometrics have potential to assess several beef physical and chemical quality traits.

Effects of the Adulteration Technique on the Near-Infrared Detection of Melamine in Milk Powder.

The United States Pharmacopeial Convention has led an international collaborative project to develop a toolbox of screening methods and reference standards for the detection of milk powder adulteration. During the development of adulterated milk powder reference standards, blending methods used to combine melamine and milk had unanticipated strong effects on the near-infrared (NIR) spectrum of melamine. The prominent absorbance band at 1468 nm of melamine was retained when it was dry-blended with skim milk powder but disappeared in wet-blended mixtures, where spray-dried milk powder samples were prepared from solution. Analyses using polarized light microscopy, Raman spectroscopy, dielectric relaxation spectroscopy, X-ray diffraction, and mass spectrometry indicated that wet blending promoted reversible and early Maillard reactions with lactose that are responsible for differences in melamine NIR spectra between wet- and dry-blended samples. Targeted detection estimates based solely on dry-blended reference standards are likely to overestimate NIR detection capabilities in wet-blended samples as a result of previously overlooked matrix effects arising from changes in melamine hydrogen-bonding status, covalent complexation with lactose, and the lower but more homogeneous melamine local concentration distribution produced in wet-blended samples. Techniques used to incorporate potential adulterants can determine the suitability of milk reference standards for use with rapid detection methods.

The application of transcriptomic data in the authentication of beef derived from contrasting production systems.

BACKGROUND: Differences between cattle production systems can influence the nutritional and sensory characteristics of beef, in particular its fatty acid (FA) composition. As beef products derived from pasture-based systems can demand a higher premium from consumers, there is a need to understand the biological characteristics of pasture produced meat and subsequently to develop methods of authentication for these products. Here, we describe an approach to authentication that focuses on differences in the transcriptomic profile of muscle from animals finished in different systems of production of practical relevance to the Irish beef industry. The objectives of this study were to identify a panel of differentially expressed (DE) genes/networks in the muscle of cattle raised outdoors on pasture compared to animals raised indoors on a concentrate based diet and to subsequently identify an optimum panel which can classify the meat based on a production system. RESULTS: A comparison of the muscle transcriptome of outdoor/pasture-fed and Indoor/concentrate-fed cattle resulted in the identification of 26 DE genes. Functional analysis of these genes identified two significant networks (1: Energy Production, Lipid Metabolism, Small Molecule Biochemistry; and 2: Lipid Metabolism, Molecular Transport, Small Molecule Biochemistry), both of which are involved in FA metabolism. The expression of selected up-regulated genes in the outdoor/pasture-fed animals correlated positively with the total n-3 FA content of the muscle. The pathway and network analysis of the DE genes indicate that peroxisome proliferator-activated receptor (PPAR) and FYN/AMPK could be implicit in the regulation of these alterations to the lipid profile. In terms of authentication, the expression profile of three DE genes (ALAD, EIF4EBP1 and NPNT) could almost completely separate the samples based on production system (95 % authentication for animals on pasture-based and 100 % for animals on concentrate- based diet) in this context. CONCLUSIONS: The majority of DE genes between muscle of the outdoor/pasture-fed and concentrate-fed cattle were related to lipid metabolism and in particular ß-oxidation. In this experiment the combined expression profiles of ALAD, EIF4EBP1 and NPNT were optimal in classifying the muscle transcriptome based on production system. Given the overall lack of comparable studies and variable concordance with those that do exist, the use of transcriptomic data in authenticating production systems requires more exploration across a range of contexts and breeds.

Dispersive Raman spectroscopy and multivariate data analysis to detect offal adulteration of thawed beefburgers.

Beef offal (i.e., kidney, liver, heart, lung) adulteration of beefburgers was studied using dispersive Raman spectroscopy and multivariate data analysis to explore the potential of these analytical tools for detection of adulterations in comminuted meat products with complex formulations. Adulterated (n = 46) and authentic (n = 36) beefburger samples were produced based on formulations derived using market knowledge and an experimental design. Raman spectral data in the fingerprint range (900-1800 cm(-1)) were examined using both a classification (partial least-squares discriminant analysis, PLS-DA) and class-modeling (soft independent modeling of class analogy, SIMCA) approach to identify offal-adulterated and authentic beefburgers. PLS-DA models correctly classified 89-100% of authentic and 90-100% of adulterated samples. SIMCA models were developed using either PCA or PLS scores as input data. For authentic beefburgers, they exhibited sensitivity, specificity, and efficiency values of 0.94-1, 0.64-1, and 0.80-0.97, respectively. PLS regression quantitative models were also developed in an attempt to quantify total offal and added fat in these samples. The performance of PLS regression quantitative models for prediction of added fat may be acceptable for screening purposes, with the most accurate model producing a coefficient of determination in prediction of 0.85 and a root-mean-square error of prediction equal to 3.8% w/w.

Characterization of near-infrared spectral variance in the authentication of skim and nonfat dry milk powder collection using ANOVA-PCA, pooled-ANOVA, and partial least-squares regression.

Forty-one samples of skim milk powder (SMP) and nonfat dry milk (NFDM) from 8 suppliers, 13 production sites, and 3 processing temperatures were analyzed by NIR diffuse reflectance spectrometry over a period of 3 days. NIR reflectance spectra (1700-2500 nm) were converted to pseudoabsorbance and examined using (a) analysis of variance-principal component analysis (ANOVA-PCA), (b) pooled-ANOVA based on data submatrices, and (c) partial least-squares regression (PLSR) coupled with pooled-ANOVA. ANOVA-PCA score plots showed clear separation of the samples with respect to milk class (SMP or NFDM), day of analysis, production site, processing temperature, and individual samples. Pooled-ANOVA provided statistical levels of significance for the separation of the averages, some of which were many orders of magnitude below 10⁻³. PLSR showed that the correlation with Certificate of Analysis (COA) concentrations varied from a weak coefficient of determination (R²) of 0.32 for moisture to moderate R² values of 0.61 for fat and 0.78 for protein for this multinational study. In this study, pooled-ANOVA was applied for the first time to PLS modeling and demonstrated that even though the calibration models may not be precise, the contribution of the protein peaks in the NIR spectra accounted for the largest proportion of the variation despite the inherent imprecision of the COA values.

Detection of adulteration in fresh and frozen beefburger products by beef offal using mid-infrared ATR spectroscopy and multivariate data analysis.

A series of authentic and offal-adulterated beefburger samples was produced. Authentic product (36 samples) comprised either only lean meat and fat (higher quality beefburgers) or lean meat, fat, rusk and water (lower quality product). Beef offal adulterants comprised heart, liver, kidney and lung. Adulterated formulations (46 samples) were produced using a D-optimal experimental design. Fresh and frozen-then-thawed samples were modelled, separately and in combination, by a classification (partial least squares discriminant analysis) and class-modelling (soft independent modelling of class analogy) approach. With the former, 100% correct classification accuracies were obtained separately for fresh and frozen-then-thawed material. Separate class-models for fresh and frozen-then-thawed samples exhibited high sensitivities (0.94 to 1.0) but lower specificities (0.33-0.80 for fresh samples and 0.41-0.87 for frozen-then-thawed samples). When fresh and frozen-then-thawed samples were modelled together, sensitivity remained 1.0 but specificity ranged from 0.29 to 0.91. Results indicate a role for this technique in monitoring beefburger compliance to label.

Simultaneous data pre-processing and SVM classification model selection based on a parallel genetic algorithm applied to spectroscopic data of olive oils.

Classification is an important task in chemometrics. For several years now, support vector machines (SVMs) have proven to be powerful for infrared spectral data classification. However such methods require optimisation of parameters in order to control the risk of overfitting and the complexity of the boundary. Furthermore, it is established that the prediction ability of classification models can be improved using pre-processing in order to remove unwanted variance in the spectra. In this paper we propose a new methodology based on genetic algorithm (GA) for the simultaneous optimisation of SVM parameters and pre-processing (GENOPT-SVM). The method has been tested for the discrimination of the geographical origin of Italian olive oil (Ligurian and non-Ligurian) on the basis of near infrared (NIR) or mid infrared (FTIR) spectra. Different classification models (PLS-DA, SVM with mean centre data, GENOPT-SVM) have been tested and statistically compared using McNemar's statistical test. For the two datasets, SVM with optimised pre-processing give models with higher accuracy than the one obtained with PLS-DA on pre-processed data. In the case of the NIR dataset, most of this accuracy improvement (86.3% compared with 82.8% for PLS-DA) occurred using only a single pre-processing step. For the FTIR dataset, three optimised pre-processing steps are required to obtain SVM model with significant accuracy improvement (82.2%) compared to the one obtained with PLS-DA (78.6%). Furthermore, this study demonstrates that even SVM models have to be developed on the basis of well-corrected spectral data in order to obtain higher classification rates.

Observations on the water distribution and extractable sugar content in carrot slices after pulsed electric field treatment.

The impact of pulsed electric field (PEF) processing conditions on the distribution of water in carrot tissue and extractability of soluble sugars from carrot slices was studied. Time domain NMR relaxometry was used to investigate the water proton mobility in PEF-treated carrot samples. Three distinct transverse relaxation peaks were observed in untreated carrots. After PEF treatment only two slightly-overlapping peaks were found; these were attributed to water present in the cytoplasm and vacuole of carrot xylem and phloem tissues. This post-treatment observation indicated an increase in water permeability of tissues and/or a loss of integrity in the tonoplast. In general, the stronger the electric field applied, the lower the area representing transverse relaxation (T2) values irrespective of treatment duration. Moreover an increase in sucrose, ß- and α-glucose and fructose concentrations of carrot slice extracts after PEF treatment suggested increases in both cell wall and vacuole permeability as a result of exposure to pulsed electric fields.

Exploring authentic skim and nonfat dry milk powder variance for the development of nontargeted adulterant detection methods using near-infrared spectroscopy and chemometrics.

A multinational collaborative team led by the U.S. Pharmacopeial Convention is currently investigating the potential of near-infrared (NIR) spectroscopy for nontargeted detection of adulterants in skim and nonfat dry milk powder. The development of a compendial method is challenged by the range of authentic or nonadulterated milk powders available worldwide. This paper investigates the sources of variance in 41 authentic bovine skim and nonfat milk powders as detected by NIR diffuse reflectance spectroscopy and chemometrics. Exploratory analysis by principal component analysis and varimax factor rotation revealed significant variance in authentic samples and highlighted outliers from a single manufacturer. Spectral preprocessing and outlier removal methods reduced ambient and measurement sources of variance, most likely linked to changes in moisture together with sampling, preparation, and presentation factors. Results indicate that significant chemical variance exists in different skim and nonfat milk powders that will likely affect the performance of adulterant detection methods by NIR spectroscopy.

Beef authentication using dietary markers: chemometric selection and modelling of significant beef biomarkers using concatenated data from multiple analytical methods.

This work aimed to identify a combination of isotopic and molecular biomarkers in bovine muscle and adipose tissue for authentication of the diet of beef cattle. Muscle and adipose tissue samples were collected from animals one of four dietary treatments fed over a 1 year period : pasture (P), barley-based concentrate (C), silage followed by pasture (SiP) or silage followed by pasture with concentrate (SiPC). In total, 83 variables were studied including volatile compounds, colour and reflectance measurements, stable isotope ratios, fatty acids, ß-carotene, lutein and α-tocopherol. Chemometric models were created for each dietary treatment using the entire and an attenuated variable set. Meat from each dietary treatment was identified with a high level of accuracy (correct classification between 90.8% and 100%) using a discriminant approach. After elimination of insignificant variables, accuracy was maintained or marginally improved. SIMCA class-modelling performed moderately well, especially with the reduced variable set (76.1-100% correct classification).

Confirmation of brand identity of a Trappist beer by mid-infrared spectroscopy coupled with multivariate data analysis.

Authentication of foods is of importance both to consumers and producers for e.g. confidence in label descriptions and brand protection, respectively. The authentication of beers has received limited attention and in most cases only small data sets were analysed. In this study, Fourier-transform infrared attenuated total reflectance (FT-IR ATR) spectroscopy was applied to a set of 267 beers (53 different brands) to confirm claimed identity for samples of a single beer brand based on their spectral profiles. Skewness-adjusted robust principal component analysis (ROBPCA) was deployed to detect outliers in the data. Subsequently, extended canonical variates analysis (ECVA) was used to reduce the dimensionality of the data while simultaneously achieving maximum class separation. Finally, the reduced data were used as inputs to various linear and non-linear classifiers. Work focused on the specific identification of Rochefort 8° (a Trappist beer) and both direct and indirect (using an hierarchical approach) identification strategies were studied. For the classification problems Rochefort vs. non-Rochefort, Rochefort 8° vs. non-Rochefort 8° and Rochefort 8° vs. Rochefort 6° and 10°, correct prediction abilities of 93.8%, 93.3% and 97.3%, respectively were achieved.

Feasibility study on the use of visible-near-infrared spectroscopy for the screening of individual and total glucosinolate contents in broccoli.

The potential of visible-near-infrared spectroscopy to determine selected individual and total glucosinolates in broccoli has been evaluated. Modified partial least-squares regression was used to develop quantitative models to predict glucosinolate contents. Both the whole spectrum and different spectral regions were separately evaluated to develop the quantitative models; in all cases the best results were obtained using the near-infrared zone between 2000 and 2498 nm. These models have been externally validated for the screening of glucoraphanin, glucobrassicin, 4-methoxyglucobrassicin, neoglucobrassicin, and total glucosinolates contents. In addition, discriminant partial least-squares was used to distinguish between two possible broccoli cultivars and showed a high degree of accuracy. In the case of the qualitative analysis, best results were obtained using the whole spectrum (i.e., 400-2498 nm) with a correct classification rate of 100% in external validation being obtained.

BAYESIAN WAVELET-BASED CURVE CLASSIFICATION VIA DISCRIMINANT ANALYSIS WITH MARKOV RANDOM TREE PRIORS.

Discriminant analysis is an effective tool for the classification of experimental units into groups. When the number of variables is much larger than the number of observations it is necessary to include a dimension reduction procedure into the inferential process. Here we present a typical example from chemometrics that deals with the classification of different types of food into species via near infrared spectroscopy. We take a nonparametric approach by modeling the functional predictors via wavelet transforms and then apply discriminant analysis in the wavelet domain. We consider a Bayesian conjugate normal discriminant model, either linear or quadratic, that avoids independence assumptions among the wavelet coefficients. We introduce latent binary indicators for the selection of the discriminatory wavelet coefficients and propose prior formulations that use Markov random tree (MRT) priors to map scale-location connections among wavelets coefficients. We conduct posterior inference via MCMC methods, we show performances on our case study on food authenticity and compare results to several other procedures..

Attempted confirmation of the provenance of Corsican PDO honey using FT-IR spectroscopy and multivariate data analysis.

This study investigated the potential of Fourier-transform infrared (FT-IR) spectroscopy and chemometric techniques to produce a mathematical model that would confirm or refute the provenance of honeys claiming to be Corsican. Authentic honey samples from two harvest seasons (2004/2005 and 2005/2006) were collected from Ireland (n=2), Italy (n=30), Austria (n=40), Germany (n=36), mainland France (n=46), and Corsica (n=219). Prior to scanning, samples were diluted with distilled water to a standard solids content (70 degrees Brix). Spectra (2500-12500 nm) were recorded at room temperature using a FT-IR spectrometer equipped with a germanium attenuated total reflectance (ATR) accessory. Standard normal variate (SNV) and first- and second-derivative data pretreatments were applied to the recorded spectra, which were processed using factorial discriminant analysis (FDA) and partial least-squares (PLS) regression analysis. Overall correct classification figures of 82% (FDA) and 87% (PLS) were obtained for a separate validation set comprising samples from both harvests.