Analysis of chicken fat as adulterant in butter using fourier transform infrared spectroscopy and chemometrics

Food adulteration is an addition process of substances which are injurious to health, or by the removal of substances which are nutritious. The driving force behind this process is profit maximization which can be achieved using lowcost substances to partially or wholly substitute with the more expensive ones (Arvaniyoyannis and Tzourous, 2005). The mixing of animal fats with food products is a major concern to certain groups of consumers due to religious obligations and health complications. From religious perspectives, the source of fat that acts as adulterant is a serious issue of concern. In Islamic and Kosher dietary laws, foods containing porcine based substances are strictly forbidden, while in Hinduism, the consumption of beef fats in food is prohibited (Eliasi and Dweyer, 2002, Marikkrar et al., 2005). Butter is undoubtedly one of the most complex of all edible fats with more than 500 different fatty acids. It is mainly comprised of saturated fatty acids (SFA), followed by monounsaturated fatty acids (MUFA), and small amounts of polyunsaturated fatty acids (PUFA). It has more than 1300 individual triacylglycerols (TAG) (Barron et al., 1990). Commercial butter must have at least 80-82% pure milk fat, water, and sometimes salt. Milk or cream should be the primary product. Butter is the foremost lipid product of animal RESUMEN


Analysis of chicken fat as adulterant in butter using fourier transform Infrared spectroscopy and chemometrics
Butter may be adulterated with cheaper animal fafs, such as chicken fat (CF). Thus, the detection and quantification of butter adulteration with CF was monitored using Fourier transform infrared (FTIR) spectroscopy, combined with chemometric of partial least square (PLS) at the frequency regions of 1200-1000cm \ FTIR measurements were made on pure butter and that adulterated with varying concentrations of CF (0-100% w/w in butter). PLS calibration exhibits a good relationship between actual and FTIR predicted values of CF with a coefficient of deteiTnination (R^) of 0.981. The root means standard error of calibration (RMSEC) and during cross validation (RMSECV) obtained using six principal components (PCs) are 2.08 and 4.33% v/v, respectively.

1.INTRODUCTION
Food adulteration is an addition process of substances which are injurious to health, or by the removal of substances which are nutritious. The driving force behind this process is profit maximization which can be achieved using lowcost substances to partially or wholly substitute with the more expensive ones (Arvaniyoyannis and Tzourous, 2005). The mixing of animal fats with food products is a major concern to certain groups of consumers due to religious obligations and health complications. From religious perspectives, the source of fat that acts as adulterant is a serious issue of concern. In Islamic and Kosher dietary laws, foods containing porcine based substances are strictly forbidden, while in Hinduism, the consumption of beef fats in food is prohibited (Eliasi andDweyer, 2002, Marikkrar etal., 2005).
Butter is undoubtedly one of the most complex of all edible fats with more than 500 different fatty acids. It is mainly comprised of saturated fatty acids (SFA), followed by monounsaturated fatty acids (MUFA), and small amounts of polyunsaturated fatty acids (PUFA). It has more than 1300 individual triacylglycerols (TAG) (Barron et al., 1990). Commercial butter must have at least 80-82% pure milk fat, water, and sometimes salt. Milk or cream should be the primary product. Butter is the foremost lipid product of animal agriculture in terms of organoleptic qualities, market price, and wide spread use in edible applications. In the U.S., the annual production of butter is slightly above 1 billion pounds (454 million kilograms, 498 million liters) (US Department of Agriculture, 2005). Due to the price difference and the similar properties, the adulteration of butter with animal faf continues to be a risk for consumers in developing countries.
Several methods have been developed for the detection and quantification of adulterants in butter. Numerous authors (De peters 1993, Carisano and Riva 1976, Coleman 1961Mattson and Lufon 1958;Mattson 1963;Jensen et al., 1964) have reported small amounts of beef tallow incorporated into butter by evaluating the fatty acid composition of the monoglycerides acquired by enzymatic hydrolysis. The addition of beef tallow in butter has been reported by Solimen and Yoones (1986) by determining the cholesterol esters and diglycerides. Currently, Precht (1991) and Mariani et al., (1990) have reported the triglycéride composition of various butter samples. The detection of 1-3% vegetable and 3-5% animal fats can be made using statistical parameters. All these methods are only applied to the natural components of fats. However, butter can also contain substances deriving from the refining processes (Lanzón et al., 1989). Therefore, the development of a rapid, accurate, inexpensive analytical technique which is capable of defecting such adulterafion in butter is pertinent and highly demanded.
Nowadays, the application of Fourier transformed infrared (FTIR) spectroscopy has emerged, mainly in food studies and has predominantly become a useful analytical tool in the study of edible fafs and oils (Guillen and Cabo, 2000). FTIR specfroscopy has received greaf attention in the quantitative analysis of fafs and oils over the years, due fo the main advantage of easy sample preparation with reduced or no-sample pre-freatment steps (Sherazi et al., 2007, Baefen andDardenne, 2002). It is used for the high-throughpuf analysis of milk-based food components that rapidly allows real-fime measurements af all stages of production without requiring special skills from users. This technique can be easily applied in fundamental research, control laboratories and industrial settings (Karoui andDe Baerdemaeker, 2007, Subramaniam and). There have been several studies concerning the classification, characterization, and aufhenficafion studies of edible fats and oils using IR specfroscopy .
In combination with prevalent chemometric techniques, FTIR spectroscopy methods have been emphasized for fhe quantitative analysis of various food products such as lard in cake formulation (Syahariza et al., 2005), biscuits (Che Man et al., 2011), cocoa buffer (Che Man ei al., 2011) and mixtures of lards with other animal fats (Che Man and Mirghani, 2001). These techniques are also proven fo assess the overall levels of butterfat and butyric acids as an indirect indicator of adulteration (Heussen et al., 2007). The adulteration of buffer faf with foreign faf could be detecfed by observing the FTIR spectra af fhe specific wavelengfh due fo fhe ratios of c/s-unsafurafions of fatty acid moiefies as reported by Safo et al., (1990). However, fhere is no information available relafed to fhe use of FTIR specfroscopy coupled wifh chemomefrics for fhe analysis of buffer adulferafed wifh chicken faf. Therefore, in fhis sfudy, we proposed FTIR specfroscopy combined wifh mulfivariafe analysis for fhe defecfion and quanfificafion of chicken faf using partial leasf square (PLS).

Sample preparation
Lard, beef, mutton and chicken were obfained through the rendering process of fhe adipose fissues of fhe corresponding animals. The rendering processed was carried out according fo Rohman and Che Man (2009a). Butter samples were exfracfed according fo fhe AOAC official mefhod 920.118 (2000). The exfracfed samples were kepf in glass vials under refrigerafed conditions (-20 °C) unfil used for analysis. Infrared specfra were collecfed for each sample fo develop a classification model.

Calibration and Validation
The calibrafion samples, composing of a number of sfandard or fraining sets consisting of chicken faf (CF) in butter af concentrafion ranges of 0-100% v/v, were prepared. For validafion, a series of independenf samples was builf fo evaluafe fhe predicfive ability of fhe developed calibrafion model. The specfra of pure butter and CF as well as fheir mixfures were analyzed using FTIR specfroscopy. The wavelengfh regions where fhe variations were observed were chosen for developing fhe PLS model in order fo quanfify CF in butter.

Statistical and chemometrics analysis
The Chemometric analysis of PLS was done using the software Horizon MB (Canada). The leave one out cross validation procedure was used to verify the calibration model. The calibration of performance of PLS was assessed using the values of root mean standard error of calibration (RMSEC) and coefficient of determination (R^). In addition, R^ and root mean standard error of prediction (RMSECV) were used for the evaluation of the validation capacity of PLS.

Fatty acid analysis
The fatty acid (FA) compositions of butter and other animal fats were determined using a gas Chromatograph (Shimadzu GC-2010, Shimadzu Corp., Tokyo, Japan), equipped with flame ionization detector. The oven temperature was programmed as follows: the initial temperature was 100°C (held for 1 min), then ramped up to 180°C (8°C min"Y increased from 180 to 240 °C (10°C min"^), and finally held at 240°C for 5 min. The temperatures of detector and injector were maintained at 240 °C during the analysis. The flow rate of carrier gas (helium) was 6.8 mL min"\ Before analysis, the samples were treated with sodium methoxyde to form FAMEs according to the method described by Chin et al., (2009). The column, oven and other conditions used during FA analysis are similar to those reported by Rohman and Che Man (2009b). The qualitative analysis of FAMEs in samples was carried out by comparing retention times of the peaks with those of FAMEs standards. The quantification of FAs was performed using the technique of internal normalization and expressed as percentages based on peak areas.

RESULTS AND DISCUSSION
The analyses focused on the measurement of the FTIR spectra of butter and chicken fat in the 4000-650 cm"^ spectral region. The characteristic infrared spectra of butter and CF are shown in Figure 1. The absorption bands of water, corresponding to the O-H groups, were observed in the region of 1600-1500 cm"\ which can affect the amide I signal at about 1650 cm"^ (Karoui and De Baerdemaker, 2007;Rodriguez et al., 2006). In agreement with Koca et al., (2010), strong absorptions were observed at 2900 and 2800 cm~\ respectively, corresponding to C-H (CHa and CHg) stretching vibrations. Moreover, a weak signal at 3000 cm"^ associated with -C=C-H stretching groups of c/s-unsaturation was observed. At 1745 cm"\ another strong band was present, which is reported to be associated with -C=O stretching vibrations of acids and esters (Lema Garcia et al., 2010). This band and the next at 1460 cm~^ arising from N-H bending vibration are most likely associated with the amide I and amide II of proteins (Karoui and De Baerdemaker, 2007;Rodriguez et al., 2006). In the last part of the spectra (1300-1000 crTi"^), stretching vibrations of the C-0 bond of esters and bending vibrations of a méthylène group were present (Lema García et al., 2010).The band at 966 cm"\ associated with -HC=CH out-of-plane deformation vibrations, has been previously reported as a marker band for the determination of frans-fats.
Butter contains more saturated fatty acids than those in CF (Table 1), especially myristic acid (C14:0), as determined using gas chromatography. Meanwhile, CF has more unsaturated fatty acids, especially linoleic acid (C18:2), compared with butter. The presence of unsaturated fatty acids in CF can also be observed in its FTIR spectrum at  3006 cm"^ which indicafes fhe higher amounf of unsafurafed fatty acids. This mefhod was developed for PLS analysis which relies on fhe exploifafion of fhese small changes in fhe regions of inferesf, namely af frequencies of 1200-1000 cm"\ Taking info accounf fhe difference between fhe butter and CF specfra, if is obvious fhaf peak infensifies af 1200-1000 cm"^ are differenf. Therefore, fhese frequencies were selecfed fo be opfimized for fhe analysis of CF in buffer, because FTIR specfra variafion was observed between CF and butter.

Quantification of chicken fat in butter
Absorbencies of CF wifh concenfrafions ranging from 0%-100% in buffer were recorded as a calibrafion model. Partial Leasf Square (PLS) was used for making a relafionship befween acfual and predicfed values of CF (%v/v) in butter. Frequencies af selecfed fingerprinf (1200-lOOOcm"^) were exploifed for quanfifafive analysis. The relafionship befween acfual value (x-axis) and fhe FTIR predicfed value of CF in fhe PLS calibrafion model is shown in Figure 2 linear regression y = 0.971 x + 1.601; was obtained with R^ and RMSEC values at 0.981 and 2.08% v/v, respectively. R^ values defined the relationship between the actual and predicted value of the analyte of interest (CF). This means that the nearer the R^ value is to unity, the better the relationship. Meanwhile, RMSEC refers to the root mean error square calibration uncertainty. The smaller the RMSEC value, the better the calibration model. The goodness of a calibration can be summarized by two values: the percent of variance explained by the model and the Root Mean Square Error in Calibration (RMSEC). The former, being a "normalized" value, gives a first idea about how much of the variance of the data set is "captured" by the model; the latter, being an absolute value to be interpreted the same way as a standard deviation is, giving information about the magnitude of error (Leardi, 2002).
The main problem in PLS algorithm is over fitting, which means that the PLS model produces a good model in the calibration dataset, but the model will not perform well in validation datasets using similar samples. In order to evaluate the over-fitting, a procedure of cross validation using the leave-one-out technique was used (Wang etal., 2006). The PLS calibration model was further subjected to cross validation using the "leave one out" technique. For the validation procedure, other samples prepared in the laboratory were used to minimize the validation error and to provide an estimate of the overall accuracy of validations. The root mean square error of cross validation (RMSECV) obtained was 4.33% v/v.
The confirmation and validation of the analysis region used for developing the PLS model were performed by computing the predicted residual error sum of squares (PRESS) values for different factors or principal components (PCs). The PRESS is a value direct measure on how well a calibration can predict the concentration left out during a cross validation (Smith, 2002), PRESS informed that the optimal factor number is 6, as revealed in Figure 3, which illustrates how the RMSEC obtain a stable value, minimally after six factors. This confirms that the spectral region used for developing the PLS model for the quantification of CF exhibits significant correlation with its concentration. From residual analysis as shown in Figure 4, it can be stated that errors occurring during analysis are random.

CONCLUSION
It can be concluded that FTIR spectroscopy in combination with chemometrics can be used to detect and to quantify the adulteration of butter and CF. The level of adulterants was successfully determined with the aid of a PLS calibration model. PLS can be successfully used to quantify the level of CF adulterant at the selected fingerprint region of (1200-1000 cm"^).
Copyright of Grasas y Aceites is the property of Instituto de la Grasa and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.