Lard adulteration in corned beef products can cause different vibration of the FTIR spectrum because lard contains a different TGA composition compared to beef

Samples of pork and beef purchased at a traditional market in Yogyakarta. Spices and other additives for the manufacture of reference samples purchased in supermarkets in the District of Umbulharjo, Yogyakarta. Materials n-hexane (p.a) (Merck), n-hexane (technical) (Merck), acetone (Merck) and anhydrous Na₂SO₄.

The corned beef was made by mixing meat, sugar, pepper, and flour. The meat was steamed for 40 minutes then packed in cans. Standard corned samples were prepared with corned beef containing 7 concentrated levels of pork in beef (Table 1).

Analysis of lard (extracted from corned beef) was carried out by spectrophotometry FTIR and processed by PLS multivariate and PCA analysis. The spectra region that showed the difference spectra of lard with other components were selected to create a model of PLS and PCA (Miller and Miller, 2010).

The accuracy of the calibration models was indicated by the RMSEC and R² value obtained from the Horizon MB software (Philadelphia, USA). While the validation models produce RMSEP, RMSECV, and R² are calculated following equation below:

(1) RMSECV =  ∑ a=1 n ( Y ^ b−Ya ) 2 N (2) RMSEP =  ∑ a=1 n ( Y ^ b−Ya ) 2 N

Where Ŷa is the actual value, Ya is the predictive value, N is the sample number, and The Ŷb is the calculated value for Ya (predictive value) based on the calibration equation with sample a (Naes, et al., 2004).

Based on the scanning, the infrared spectra showed that both pork and beef corned products contain different vibration of functional groups. The vibration C = O as the ester group was shown at wavenumber 1747 cm^-1. These peaks arise because the fat structure is a triglyceride consisting of the carbonyl ester group. Besides, the stretching vibration C – O group was identified at wave number 1238 cm^-1 for beef, while pork has stretching vibration C – O group at 1234 cm^-1. The indication ester vibration was also illustrated at wavenumbers 1157 cm^-1 and 1099 cm^-1 (Jaswir et al., 2003).

The vibration of the alkenes group (C = C) was illustrated at wavenumbers 3008 cm^-1 and 1654 cm^-1. The different spectra produced between both pork and beef corned products showed that they have a different composition (Figure 1). It can be said pork contains more unsaturated fatty acids level compared with beef. Belitz, Grosch, and Schieberle (2009) reported that pork contains high unsaturated fatty acids (double bonds) including oleic acid (43%) and linoleic acid (9%), while beef contains less unsaturated fatty acid than pork. Detailed vibration of all functional groups is given in Table 2.

However, quantitative analysis of lard in the corned products showed that the optimum difference of both pork and beef corned products was obtained at the wavenumber range of 1180 – 730 cm^-1. For meat discriminant, pork, beef, and mutton can be seen at 2925 cm⁻¹, 2855 cm⁻¹, and 1745 cm⁻¹ with strong peaks and weak peaks at 750 cm⁻¹ and 1800 cm⁻¹ as fingerprint regions (Yang et al., 2017). Besides, lard detection in cake formulation can be performed by FTIR at 1117 – 1097 and 990 – 950 cm⁻¹ (Syahariza et al., 2005) and lard identification in steamed sausages and grilled sausages at wavenumbers of 1000 – 791 cm^-1 and 1070 – 796 cm^-1, respectively (Guntarti, Ahda, and Sunengsih, 2019). Even, lard adulteration in palm oils can be detected at 3006 and 1117 cm⁻¹ (Sim, Chai and Kimura, 2018) or 3006 cm⁻¹, and 1120 – 1095 cm⁻¹ (Che Man, et al., 2013) or 1481.22 – 999.05 cm⁻¹ and 1793.67 – 1650.95 cm⁻¹ (Ahda and Safitri, 2016). Therefore, each sample type will change a marker region for lard detection. Based on the wavenumber range of 1180 – 730 cm^-1, the result of regression equation is y = 0.95x + 1.32, with R² value of 0.983 and RMSEC value of 2.055. The calibration equation is also validated using external validation and cross-validation. The results of crossvalidation curve y = 0.9721x - 0.5334 with R² values 0.9995 and RMSECV value 2.22%. Besides, external validation has R² values of 0.9984, RMSEP of 1.65%, with the equation y = 0.9940x - 0.7061. the higher R² value and lower error (RMSEC and RMSECV) are the good indication of the obtained equation (Rohman et al., 2012^b; Rohman, Setyaningrum, and Riyanto, 2014; Ahda et al., 2020).

Based on external validation and cross-validation were indicated that the calibration model at wave number 1180 – 730 cm^-1 was able to give accurate results for the quantitative analysis of pork in the corned mixtures.

Analysis of pork in a food product can be performed and grouped by PCA analysis. It is an analytical technique to reduce the data when it found a correlation between the data (Garcia, 2012). The PCA is unsupervised pattern recognition techniques widely used for the classification of different samples (Nunes, 2014; Rahmania, Sudjadi and Rohman, 2015). PCA analysis will reduce the number of independent variables in the data to produce the new variables that are called the principal component or major components (Che Man, Syahariza, and Rohman, 2010). Hence, the wavenumber regions for PCA were also optimized. Finally, the same wavenumbers used for quantitative analysis were chosen for PCA modeling due to its capability to provide good separation among the evaluated samples (Rahayu et al., 2018; Gamperline, 2006).

In this study, pork analysis in the corned product is performed in optimum condition at wavenumbers of 1180- 730 cm^-1. The discriminant analysis showed that 100% pork corned and 100% beef corned can be separated and distinguished (Lumakso et al., 2015), it illustrated that both corned products have a different composition (Figure 1). Therefore, analysis of pork adulteration in the commercial corned products can be performed in similar conditions (Van der Spiegel et al., 2012).

In this research, we identify 3 commercial corned products and also observe the cooking effect in disrupting halal analysis. The result showed that 3 commercial corned products are not at all produced from beef. Sample (f) are grouped as pork corned, hence we can estimate that it is made from pork (Figure 2). However, all samples are grouped in the beef corned product after the cooking process. It showed that the cooking process can affect the chemical properties of pork. The unsaturated fatty acids of pork may degrade during the heating process because pork corned is not separated from beef corned (Bhaskar et al., 2012). Therefore, halal authentication using FTIR combined with chemometrics has a problem if the product is carried out by different process because they are possible to degrade during the process (El-Gindy, Emara, and Mostafa, 2006).