Fourier transform infrared spectroscopy and multivariate analysis for the detection and quantification of different milk species

The authenticity of milk and milk products is important and has extended health, cultural, and financial implications. Current analytical methods for the detection of milk adulteration are slow, laborious, and therefore impractical for use in routine milk screening by the dairy industry. Fourier transform infrared (FT-IR) spectroscopy is a rapid biochemical fingerprinting technique that could be used to reduce this sample analysis period significantly. To test this hypothesis we investigated 3 types of milk: cow, goat, and sheep milk. From these, 4 mixtures were prepared. The first 3 were binary mixtures of sheep and cow milk, goat and cow milk, or sheep and goat milk; in all mixtures the mixtures contained between 0 and 100% of each milk in increments of 5%. The fourth combination was a tertiary mixture containing sheep, cow, and goat milk also in increments of 5%. Analysis by FT-IR spectroscopy in combination with multivariate statistical methods, including partial least squares (PLS) regression and nonlinear kernel partial least squares (KPLS) regression, were used for multivariate calibration to quantify the different levels of adulterated milk. The FT-IR spectra showed a reasonably good predictive value for the binary mixtures, with an error level of 6.5 to 8% when analyzed using PLS. The results improved and excellent predictions were achieved (only 4-6% error) when KPLS was employed. Excellent predictions were achieved by both PLS and KPLS with errors of 3.4 to 4.9% and 3.9 to 6.4%, respectively, when the tertiary mixtures were analyzed. We believe that these results show that FT-IR spectroscopy has excellent potential for use in the dairy industry as a rapid method of detection and quantification in milk adulteration.     

用遗传算法提取南疆红枣总糖的近红外光谱特征波长

本研究尝试利用近红外光谱技术测量红枣的总糖含量,针对采用偏最小二乘（PLS）法建立近红外光谱预测模型时波长筛选问题,提出用联合区间偏最小二乘法（si PLS）与遗传算法（GA）相结合的方法遗传联合区间偏最小二乘法（GA-si PLS）来提取近红外光谱特征区域和特征波长,提高模型预测精度的方法。结果表明:将全谱等分成20个子区间,用联合区间偏最小二乘法优选出4个特征子区间,在这4个子区间的基础上再用遗传偏最小二乘法继续筛选出12个特征波长。用12个特征波长建立的偏最小二成模型精度要好于全谱建立的模型,其主因子数减少了4个,预测集标准偏差（RMSECP）减少了25%,预测相关系数（RP）提高了5%。该方法选取的波长变量建立的校正模型,不仅使模型简洁、优化,而且增强了模型的预测能力。      

Simultaneous Determination of Danofloxacin and Flumequine in Milk Based on Fluorescence Spectroscopy and Chemometrics Tools

The aim of this paper was to develop a rapid screening method to determine danofloxacin (DANO) and flumequine (FLU) in milk by fluorescence spectroscopy combined with three different chemometric tools. In this study, 2-D fluorescence data and multivariate calibration based on a partial least squares discriminant analysis (PLS-DA) regression were combined to simultaneously qualify and quantify DANO and FLU concentrations in commercial ultra-high-temperature (UHT) sterilized and pasteurized milk. Calibration sets based on the UHT whole milk from brand A were built and performed using a partial least squares (PLS) regression after deproteinization. Prediction sets based on 13 types of milk were analyzed using principal component analysis (PCA), principal PLS-DA, and PLS regression models. The multivariate calibration models were better able to determine the DANO and FLU concentrations than the univariate models, and these models could be applied to other types of milk. In contrast to the PLS-DA, which had good sensitivity and specificity, the PCA yielded less satisfactory results. In the quantitative analysis, the recoveries of the two analytes were reasonable and the root mean square error of prediction was within the acceptable range. The relative standard deviations of the predicted DANO and FLU concentrations on the various testing days were 9.2 and 6.2 %, respectively, demonstrating that the analytical method had a good reproducibility.     

Use of MALDI-TOF MS technology to evaluate adulteration of small ruminant milk with raw bovine milk

Detection of adulteration of small ruminant milk is very important for health and commercial reasons. New analytical and cost-effective methods need to be developed to detect new adulteration practices. In this work, we aimed to explore the ability of the MALDI-TOF mass spectrometry to detect bovine milk in caprine and ovine milk using samples from 18 dairy farms. Different levels of adulteration (0.5, 1, 5, 10, 20, 40, 60, and 80%) were analyzed during the lactation period of goat and sheep (in May, from 60 to 90 d in milk, and in August, from 150 to 180 d in milk). Two different ranges of peptide-protein spectra (500–4,000 Da; 4–20 kDa) were used to establish a calibration model for predicting the concentration of adulterant using partial least squares and generalized linear model with lasso regularization. The low molecular weight part of the spectra together with the generalized linear model with lasso regularization regression model appeared to have greater potential for our aim of detection of adulteration of small ruminants' milk. The subsequent prediction model was able to predict the concentration of bovine milk in caprine milk with a root mean square error of 11.4 and 17.0% in ovine milk. The results offer compelling evidence that MALDI-TOF can detect the adulteration of small ruminants' milk. However, the method is severely limited by (1) the complexity of the milk proteome resulting from the adulteration technique, (2) the potential degradation of thermolabile proteins, and (3) the genetic variability of tested samples. Additionally, the root mean square error of prediction based only on one individual sample adulteration series can drop down to 6.34% for quantification of adulterated caprine milk and 6.28% for adulterated ovine milk for the full set of concentrations or down to 2.33 and 4.00%, respectively, if we restrict only to low concentrations of adulteration (0, 0.5, 1, 5, 10%).     

The Effect of Path Length on the Measurement Accuracies of Wine Chemical Parameters by UV,Visible, and Near-Infrared Spectroscopy

The use of spectral measurements using either UV, visible (VIS), or near-infrared (NIR) spectroscopy to characterize wines or to predict wine chemical composition has been extensively reported. However, little is known about the effect of path length on the UV, VIS, and NIR spectrum of wine and the subsequent effect on the performance of calibrations used to measure chemical composition. Several parameters influence the spectra of organic molecules in the NIR region, with path length and temperature being one of the most important factors affecting the intensity of the absorptions. In this study, the effect of path length on the standard error of UV, VIS, and NIR calibration models to predict phenolic compounds was evaluated. Nineteen red and 13 white wines were analyzed in the UV, VIS, and NIR regions (200–2500 nm) in transmission mode using two effective path lengths 0.1 and 1 mm. Principal component analysis (PCA) and partial least squares (PLS) regression models were developed using full cross validation (leave-one-out). These models were used to interpret the spectra and to develop calibrations for phenolic compounds. These results indicated that path length has an effect on the standard error of cross validation (SECV) absolute values obtained for the PLS calibration models used to predict phenolic compounds in both red and white wines. However, no statistically significant differences were observed (p > 0.05). The practical implication of this study was that the path length of scanning for wines has an effect on the calibration accuracies; however, they are non-statistically different. Main differences were observed in the PCA score plot. Overall, well-defined protocols need to be defined for routine use of these methods in research and by the industry.     

Evaluation of Designed Calibration Samples for Casein Calibration in Fourier Transform Infrared Analysis of Milk

The determination of casein in milk by infrared spectrometry still suffers from deficiencies specific to the spectrometers used. Measurements with filtometers are based on the difference in the absorbance before and after casein precipitation. Computerised Fourier Transform Infrared (FT-IR) spectrometry allows the use of more spectral information but robust casein calibrations were achieved only with high numbers of natural calibration samples. This is expensive due to the costly reference analyses needed for calibration. To reduce effort and costs of casein calibrations we studied the usefulness of commercial calibration milks (designed for fat, protein, and lactose calibration) for casein calibrations. We tested calibration models with and without natural milk samples and assessed the validation results according to International Organization for Standardization and International Dairy Federation standards. We found the combined use of commercial designed samples and fresh natural raw milk samples a very promising approach to the partial least squares casein calibration of FT-IR spectrometers. Measurement uncertainties for casein as low as 0.020 g/100 g for the bias and 0.047 g/100 g for the standard error of prediction (on 25 validation samples of high variability) were achieved with a calibration of 38 samples. This accuracy complies with the requirements of the International Dairy Federation standard 141C:2000 for the infrared analysis of milk.     

Determination of Protein Content of Raw Fresh Cow’s Milk Using Dielectric Spectroscopy Combined with Chemometric Methods

To investigate the feasibility of dielectric spectroscopy in determining the protein content of raw fresh milk, the dielectric spectra of dielectric constant and loss factor were obtained on 145 raw cow’s milk samples at 201 discrete frequencies from 20 to 4500 MHz using a network analyzer and an open-ended coaxial-line probe. It was found that in below 1000 MHz, there was positive linear relationship between the loss factor and the protein content with a coefficient of determination (R ²) greater than 0.66. In order to identify the most accurate method for determining the protein content, 97 milk samples were selected for calibration set, and the other 48 samples were used as prediction set by using joint x–y distance sample set partitioning method. The standard normal variate method was used to preprocess spectra. Ten, 152, and 7 variables were extracted as characteristic variables using successive projection algorithm (SPA), uninformative variable elimination (UVE) method based on partial least squares, and SPA after UVE (UVE-SPA) methods, respectively. The results showed that applying SPA after UVE was helpful to extract indispensible characteristic variables from full dielectric spectra. The models based on the least squares supporting vector machine (LSSVM) offered the best performance at the same characteristic variable extraction method when compared with those established by partial least squares regression and extreme learning machine. The best model for determining the protein content of milk was LSSVM-UVE-SPA with \( {R}_p^2 \) of 0.865, root-mean-square error of prediction set (RMSEP) of 0.094, and residual predictive deviation (RPD) of 2.604. The results indicate that the protein content of milk could be determined precisely by using dielectric spectroscopy combined with chemometric methods. The study is helpful to explore a new milk protein sensor which could be used in situ or online measurement.     

On-line versus off-line NIRS analysis of intact olives

Visible/near-infrared calibrations were developed for the determination of the quality parameters (fat content, moisture and free acidity) of intact olive fruits. The reflectance spectra were acquired in two different instruments (diode-array versus grating monochromator based instruments). The grating monochromator based instrument was used at the laboratory (off-line analysis), whereas the portable diode-array based device was placed on top of a conveyor belt set to simulate measurements in an olive oil mill plant (on-line analysis). Partial least squares (PLS) regression and least squares support vector machine (LS-SVM) were used for the development of the calibration models. A total of 174 samples were prepared for the calibration (N = 122) and validation (N = 52) sets. The root mean square error of prediction (RMSEP) and the residual predictive deviation (RPD) values were better using the diode-array instrument and applying the PLS regression method for the fat content parameter while for the free acidity and moisture content, the LS-SVM algorithm gave the best results. The results obtained seems to suggest the viability of the on-line system, instead of the off-line analysis, for the determination of physicochemical composition in intact olives.     

基于近红外光谱技术的海参品质快速检测

目的应用近红外光谱技术建立海参产地区分和胶原蛋白快速检测的方法。方法总计43个海参样品来自大连、福建、连云港、山东4个地区。首先采集样品的近红外光谱图,经过标准正态变量(standard normal variables,SNV)预处理,利用不同定性判别模型对海参产地进行区分。通过分光光度计法测定海参的胶原蛋白含量,利用偏最小二乘法(partial least squares,PLS)、区间偏最小二乘法(interval partial least squares,iPLS)、向后区间偏最小二乘法(backwards interval partial least squares,BiPLS)和联合区间偏最小二乘法(synergy interval partial least squares,Si PLS)建立了海参胶原蛋白含量的预测模型。结果产地区分模型中最小二乘支持向量机(least-squares support vector machine regression,LS-SVM)的识别率最高,校正集识别率为100%,预测集识别率为95.35%;海参胶原蛋白预测模型中BiPLS的预测效果较好,校正集相关系数Rc为0.9002,预测集相关系数Rp为0.8517。结论近红外光谱技术可实现对海参的产地区分和胶原蛋白的快速检测。     

Prediction of fatty acid profiles in cow,ewe, and goat milk by mid-infrared spectrometry

Mid-infrared (MIR) spectrometry was used to estimate the fatty acid (FA) composition in cow, ewe, and goat milk. The objectives were to compare different statistical approaches with wavelength selection to predict the milk FA composition from MIR spectra, and to develop equations for FA in cow, goat, and ewe milk. In total, a set of 349 cow milk samples, 200 ewe milk samples, and 332 goat milk samples were both analyzed by MIR and by gas chromatography, the reference method. A broad FA variability was ensured by using milk from different breeds and feeding systems. The methods studied were partial least squares regression (PLS), first-derivative pretreatment + PLS, genetic algorithm + PLS, wavelets + PLS, least absolute shrinkage and selection operator method (LASSO), and elastic net. The best results were obtained with PLS, genetic algorithm + PLS and first derivative + PLS. The residual standard deviation and the coefficient of determination in external validation were used to characterize the equations and to retain the best for each FA in each species. In all cases, the predictions were of better quality for FA found at medium to high concentrations (i.e., for saturated FA and some monounsaturated FA with a coefficient of determination in external validation >0.90). The conversion of the FA expressed in grams per 100 mL of milk to grams per 100 g of FA was possible with a small loss of accuracy for some FA.     

A rapid method to quantify casein in fluid milk by front-face fluorescence spectroscopy combined with chemometrics

Casein in fluid milk determines cheese yield and affects cheese quality. Traditional methods of measuring casein in milk involve lengthy sample preparations with labor-intensive nitrogen-based protein quantifications. The objective of this study was to quantify casein in fluid milk with different casein-to-crude-protein ratios using front-face fluorescence spectroscopy (FFFS) and chemometrics. We constructed calibration samples by mixing microfiltration and ultrafiltration retentate and permeate in different ratios to obtain different casein concentrations and casein-to-crude-protein ratios. We developed partial least squares regression and elastic net regression models for casein prediction in fluid milk using FFFS tryptophan emission spectra and reference casein contents. We used a set of 20 validation samples (including raw, skim, and ultrafiltered milk) to optimize and validate model performance. We externally tested another independent set of 20 test samples (including raw, skim, and ultrafiltered milk) by root mean square error of prediction (RMSEP), residual prediction deviation (RPD), and relative prediction error (RPE). The RMSEP for casein content quantification in raw, skim, and ultrafiltered milk ranged from 0.12 to 0.13%, and the RPD ranged from 3.2 to 3.4. The externally validated error of prediction was comparable to the existing rapid method and showed practical model performance for quality-control purposes. This FFFS-based method can be implemented as a routine quality-control tool in the dairy industry, providing rapid quantification of casein content in fluid milk intended for cheese manufacturing.     

近红外光谱结合区间偏最小二乘法应用于花生油酸价的测定

采用近红外光谱法结合不同区间偏最小二乘波长筛选法建立花生油酸价的定量分析模型。采用酸碱滴定法测定花生油样本的酸价同时采集近红外光谱数据;采用区间偏最小二乘法(iPLS)、向后区间偏最小二乘法(BiPLS)、移动窗口偏最小二乘法(mwPLS)优选光谱特征区间;采用偏最小二乘法(PLS)对优选出来的谱段建立酸价的定量模型。结果表明,采用mwPLS选择的谱段建立的模型预测效果最佳,RMSECV和RMSEP分别为0.247 76和0.131 5,校正相关系数和预测相关系数分别为0.993 2和0.996 9。因此,近红外光谱结合移动窗口偏最小二乘法可以快速准确测定花生油的酸价。     

A Novel Technique on Determining Water Content in Milk Using Radio-Frequency/Microwave Dielectric Spectroscopy and Chemometrics

Water is the major component of milk. More water means less total solids concentration. Therefore, routine analysis on water content is very important in milk factory. To offer a new method for determining water content of milk in situ and in-line quality monitoring, dielectric spectroscopy, an instrumental method used to obtain spectra describing dielectric properties of materials, was used to determine the water content of milk in this study. The dielectric spectra of 161 milk samples were obtained at 201 discrete frequencies on a logarithmic scale from 20 to 4500 MHz at 25 °C. Ten, 34, and 14 optimal dielectric variables (ODVs) were extracted from the full dielectric spectra (FDS) with 402 dielectric variables by using successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), and combination of CARS and SPA (CARS + SPA), respectively. Four models, including two linear models, i.e., multiple linear regression (MLR) and partial least squares regression (PLSR), and another two artificial neural network models, i.e., extreme learning machine (ELM) and least squares-support vector machine (LSSVM), were established. The artificial neural network models had better prediction performance than the linear models at the same ODV selection methods. Moreover, the ODVs extracted by CARS could give better prediction performance than SPA and CARS + SPA. Among all developed models, the FDS-LSSVM model had the lowest root mean squares error of prediction set (0.054%), followed by CARS-LSSVM with 0.094%. Few variables and high prediction precision of CARS-LSSVM have great potential in developing portable milk water detector used in situ and in-line inspection. This study indicated that dielectric spectroscopy is a promising approach for accurately determining the water content of milk.     

Development of Fourier-transformed mid-infrared spectroscopy prediction models for major constituents of fractions of delactosated,defatted milk obtained through ultra- and nanofiltration

Milk filtration procedures are gaining relevance in the dairy industry because milk ultra- and nanofiltrates are used to increase milk processing efficiency, and as additives for products with improved nutraceutical properties. This study aimed to develop Fourier-transformed mid-infrared spectroscopy calibrations for ultra- and nanopermeate and retentate fractions of defatted and delactosated milk. A total of 154 samples from different milk fractions were collected and analyzed using reference methods to determine protein, solids-not-fat, glucose, and galactose content. The obtained values were matched with their respective Fourier-transformed mid-infrared spectroscopy spectra to develop new prediction models. Calibrations for each trait were built following 3 different approaches to get the best prediction models: (1) using the entire data set, (2) using 3 subsets based on component concentrations (level approach), and (3) using hierarchical clusters calculated with pairwise Mahalanobis distance among spectra (cluster approach). Calibrations were developed using partial least squares regression, after removing low signal-to-noise ratio wavelengths, and validated through a leave-one-out cross-validation procedure. In addition, the accuracy of the predicted values within each fraction was checked for each approach. Dividing the data set into subsets improved prediction models for each trait and for the samples in each milk fraction. Without considering milk fraction, the best improvement was observed for glucose and galactose. Glucose ratio performance deviation in cross-validation (RPD) increased from 7.42 to 11.31 and 11.06, for cluster and level approaches, respectively, whereas galactose RPD increased from 8.86 to 11.69 and 11.27 for cluster and level approaches, respectively. Considering milk fractions, the best improvement was observed for protein content, where RPD ranged from 0.08 to 6.06 for the whole data set calibration, whereas it ranged from 0.43 to 40.34 for the subset calibration approaches. Cluster and level approaches to build calibration models were comparable for samples from different fractions, suggesting that the 2 subsetting protocols should be both investigated to get the best prediction performances.     

基于近红外光谱对牛奶中掺杂尿素的判别分析

采集40个合格的纯牛奶样品,并配制含有尿素为1～20g/L的40个牛奶样品,研究掺杂尿素牛奶的二维相关近红外特性,在此基础上选择波数4200～4800cm-1为建模区间,采用偏最小二乘法建立定性、定量模型。结果指出通过判别偏最小二乘法可以实现纯牛奶及掺杂尿素牛奶的定性鉴别,判别正确率为100%;掺杂牛奶校正集相关系数R为0.999,交叉验证均方差为0.242,对未知样品集预测相关系数R达到0.999,预测标准偏差为0.57,这表明所建模型具有较好的预测效果。     

Near‐infrared spectroscopy and data analysis for predicting milk powder quality attributes

Near‐infrared (NIR) spectroscopy is a rapid analytical method for food products. In this study, NIR spectroscopy, data pretreatment techniques and multivariate data analysis were used to predict fine particle size fraction, dispersibility and bulk density of various milk powder samples, which are believed to have a significant impact on milk powder quality. Predictive models using partial least‐squares (PLS) regression were developed using NIR spectra and milk powder physical and functional properties, and it was concluded that the PLS models predicted milk powder quality with an accuracy of 88‐90 per cent.     

A novel method for the quantification of bovine colostral immunoglobulin G using infrared spectroscopy

Transmission infrared (IR) spectroscopy, in combination with partial least squares (PLS) regression, was used as the basis to develop a new method to quantify IgG in bovine colostrum. Colostrum samples (n = 250) were tested simultaneously by the reference radial immunodiffusion (RID) assay and IR spectroscopy. Colostral IgG concentrations obtained by RID assay were linked to pre-processed spectra and divided into two sets, i.e., calibration and test. PLS regression was applied to the calibration set and calibration models were developed, and the test set was used to assess the accuracy of the analytical method. The Pearson and concordance correlations between test set IgG concentrations as determined by the IR assay and the RID assay were 0.91. The Bland–Altman plot showed no evidence of systematic bias between IR and RID methods. Transmission IR spectroscopy is an effective method for quantification of bovine colostral IgG concentration and for assessment of colostrum quality.     

高光谱成像技术对灵武长枣果皮强度的无损检测

利用高光谱技术对灵武长枣果皮强度检测进行研究,为灵武长枣外部品质无损检测提供科学方法。采集120个灵武长枣的400¹000 nm的高光谱图像,对光谱数据进行预处理;应用连续投影算法（SPA）、正自适应加权算法（CARS）和无信息变量消除法（UVE）对原始光谱数据提取特征波长;分别建立基于全光谱和特征波长的偏最小二乘回归（PLSR）和最小二乘支持向量机（LS-SVM）果皮强度预测模型。结果表明:采用标准正态变换（SNV）预处理算法效果最优,其PLSR模型的交叉验证相关系数（Rcv）为0.8207,交叉验证均方根误差（RMSECV）为9.9630;利用SPA、CARS和UVE法从全光谱的125个波长中分别选取出29个、31个和31个特征波长;而基于全光谱建立的LS-SVM模型效果最优,其预测相关系数（Rp）为0.9555,预测均方根误差（RMSEP）为3.8282;研究结果表明基于高光谱成像技术采集的灵武长枣漫反射光谱进行果皮强度无损检测具有可行性。      

Effect of homogenisation on detection of milk protein content based on NIR diffuse reflectance spectroscopy

Investigating the effect of homogenisation on the prediction performance of protein content by using near-infrared (NIR) spectroscopy is helpful to improve protein determination precision. For this purpose, the influence of homogenisation on milk fat globules and NIR spectra was analysed firstly. Then, NIR spectra of eighty-seven cow milk samples before and after homogenisation were obtained. Multiplicative scatter correction was used to do spectral pretreatment. Uninformative variable elimination based on partial least squares (UVE-PLS) and successive projection algorithm was used to extract characteristic variables. Partial least squares regression (PLSR) and least squares support vector machine models were established. The results showed that homogenisation made milk fat globules be distributed evenly, decreased the size of fat globules and improved NIR spectral repeatability and prediction precision on protein content. The best model was PLSR-UVE-PLS, having good and excellent protein prediction ability for un-homogenised milk (RMSEP = 0.06 g/100 g, RPD = 2.69) and homogenised milk (RMSEP = 0.04 g/100 g, RPD = 3.59), respectively.     

Transfer of near infrared spectrometric models for silage crude protein detection between different instruments

This study was undertaken to distinguish different collected spectra from near infrared reflectance spectroscopy (NIRS) for silages using different configurations and types of NIRS, and how well the different techniques for transferring NIRS calibrations perform for silage crude protein detection. In the study, 2 Fourier transform instruments and 1 scanning grating instrument were involved. Five correction and transfer methods were tested and evaluated: slope/bias, local centering, orthogonal signal correction, direct standardization, and piecewise direct standardization. We concluded that the spectra obtained with 3 instruments were different and not solely due to the differences in offset. All of the methods for calibration transferring between 2 Fourier transform instruments and 1 Fourier transform instrument versus 1 scanning grating instrument could improve the predictions, but not all of the results could be accepted. The slope/bias, orthogonal signal correction, and local centering techniques were successful for calibration transferring of 2 Fourier transform instruments, considering their good performance. The best result was given for orthogonal signal correction ahead of the other 4 techniques for transferring calibrations between instruments of Fourier transform and scanning grating, and it was evaluated as moderately useful.