Methanol, acetonitrile, water and isopropanol were all LC-MS grade and obtained from Thermo Fisher Scientific (Fairlawn, NJ, USA). Ammonium format, formic acid and HPLC-grade methanol were purchased from E. Merck (Darmstadt, Germany). A calibrant solution of Pierce LTQ Velos positive and Pierce negative was obtained from Thermo Fisher Scientific (Rockford, IL, USA).

Lard, chicken fat, and beef tallow were obtained from the rendering of corresponding animals’ adipose tissues according to Rohman & Che Man (2010). An amount of 20 mg fat sample was weighed and placed in a 2-mL microcentrifuge tube. Samples of pure lard, pure chicken fat, and pure beef tallow were prepared. The adulterated beef tallow and chicken fat with lard were prepared by mixing beef tallow and chicken fat with lard using a ratio of 50:50 (% w/w). It was aimed to observe the profile of beef tallow and chicken fat when the adulteration was present. The ratio (50:50) was chosen because adulteration is usually performed in high concentrations. Each sample was dissolved in 1 mL isopropanol, then vortexed for 1 min at room temperature. Subsequently, the sample was ultrasonicated at room temperature for 30 min. After sonication finished, the sample was then centrifuged at 12,000 x g for 10 min at 4 ^oC. The supernatant was collected and filtered using PTFE filter 0.22 µm and placed into a clear HPLC vial for lipidomic analysis using LC-HRMS. Each sample was prepared in three replicates.

The lipidomic analysis was performed using an ultra-high-performance liquid chromatography (Thermo Scientific^TM Vanquish^TM UHPLC binary pump) and high-resolution mass spectrometry-Orbitrap (Thermo Scientific^TM Q-Exactive^TM Hybrid Quadrupole-Orbitrap^TM High Resolution Mass Spectrometer). The separation of analyte was carried out using an analytical column of Thermo Scientific^TM Accucore^TM C-18 (100 mm x 2.1 mm ID x 2.6 µm). Analysis was performed according to Jia et al. (2022) with modifications. Lipidomic analysis was performed using a mobile phase of water:acetonitrile (40:60 v/v) containing 40 mM ammonium format and 0.1% formic acid as the mobile phase A and isopropanol:acetonitrile (90:10 v/v) containing 40 mM ammonium format and 0.1% formic acid as the mobile phase B. The gradient mode was applied as follows: initially, the mobile phase B was set at 32% B for 1.5 min, then increased to 45% B until reaching a minimum of 4.0. After that, it was increased to 54% B (4.01-5.0 min), 58% B (5.01-8.0 min), 66% B (8.01-11 min), 70% B (11.01-14.00 min), 75% B (14.01-18.00 min), 97% B (18.01-21.00 min), then held at 97% B for 25 min. At the end, the process was returned to its initial condition (32% B) for 25.01-30.00 min. The flow rate of the mobile phase was 0.260 mL/min with a sample injection volume of 5 µL. The temperature of the sampler was set at 25 ^oC, while the column temperature was maintained at 40 ^oC. The mass spectrometry condition for untargeted lipidomic screening was carried out using full MS/dd-MS2 acquisition mode. Lipid analysis was performed both in positive and negative ionization modes. The sheath gas flow rate, auxiliary gas flow rate, and sweep gas flow rate applied in this research were set at 32, 8, and 4 arbitrary unit (AU), respectively. The electrospray ionization used spray voltage of 3.30 kV with capillary temperature set at 320 ^oC. The auxiliary gas heater temperature was set at 30 ^oC. The analysis was performed using a scan range of 100-1500 m/z and a resolution of 70,000 for full MS and 17,500 for dd-MS2. The mass spectrometer instrument was weekly calibrated using Thermo Scientific Pierce ESI calibration solutions both in positive and negative modes to warrant the mass accuracy.

The raw data of the total ion chromatogram (TIC) obtained from the LC-HRMS measurement both in positive and negative ionization modes were analyzed using Lipid Search 4.2 software (Thermo Scientific, USA) for peak alignment, baseline correction, background correction, retention time alignment (0.2 min tolerance) and mass tolerance (5 ppm). The identification of the lipid compositions was compared to the predicted in silico spectra from various lipid compounds. The results of lipid metabolomes were classified according to their lipid groups and lipid ions. Data were filtered using RSD (relative standard deviation) < 20 and S/N ratio > 10. The molecules with RSD at more than 30% and missing values exceding 50% were deleted.

Chemometrics was carried out using variables of lipid ions and the relative areas. Analysis was performed using SIMCA 14.0 software (Umetrics, Sweden). Principal component analysis (PCA) and partial least square-discriminant analysis (PLS-DA) were used in this study. The PCA model was evaluated using PCA score plot, R² value and Q² value. In addition, the PLS-DA model was evaluated using PLS-DA score plot, R²X, R²Y, and Q² values. The permutation test using 999 permutations and receiver operating characteristics (ROC) value were used to validate the PLS-DA model. The identification of potential biomarkers which are important for sample discrimination was performed using the variable importance for projections (VIP) value in the PLS-DA analysis. Variables with a VIP value higher than 1 were considered as discriminating metabolites which are potential for biomarkers.

The physical appearance of lard, beef tallow (BT), and chicken fat (CF) are similar, thus making them vulnerable for adulteration and mislabelling. Figure 1 shows the total ion chromatogram (TIC) of lard, BT, and CF obtained from the LC-Orbitrap HRMS measurement. The TIC of those three samples were very similar, thus it is very difficult to differentiate lard, BT, and CF only by using visual observation on the TIC. The lipid compositions of lard, BT, and CF were successfully identified using Lipid Search software by extracting the raw TIC data. A total of 281 lipid ions from 18 lipid groups was obtained in lard using the positive ionization mode and 11 lipid ions from two lipid groups were observed using the negative ionization mode. The main lipid composition of lard was triglycerides (TG = 51.03%) followed by diglycerides (DG = 19.52%) and ceramides (Cer = 8.90%). BT contained 339 lipid compounds from 12 lipid groups observed in the positive ionization mode as well as 6 lipid compounds from the negative ionization mode. The mos abundant lipid compositions in BT were TG (53.33%), DG (25.80%), and cer (8.99%), respectively. In addition, the main compositions of lipids in CF were also the same as lard and BT, which were TG (58.31%), DG (25.31%), and cer (5.21%), respectively. The total lipid compounds observed in CF were 395 compounds from the positive ionization mode and 8 compounds from the negative ionization mode.

Many lipid compounds in lard, BT, and CF could be found from various lipid groups. Figure 2 illustrates the Venn diagram of the lipid metabolites contained in the three types of fats. According to the diagram, it could be observed that 117 lipids were found only in lard while 172 and 210 lipids were only found in BT, and CF, respectively. On the other hand, 82 lipid compounds were identified in all three types of fats (lard, CF and BT). Further investigation detected specific lipid groups only found in lard such as AcHexStE (acyl hexosyl stigmasterol ester), BiotinylPE (biotinyl phosphoetanolamine), LPC (lysophosphatidylcholine), MePC (monoether phosphatidylcholine), PC (phosphatidylcholine) and PI (phosphoinocitides). These lipid groups were absent from CF and BT. Lipid groups of SPH (sphingomyelin) and WE (wax esters) were found to be specific to CF, whereas lipid groups of MG (monoglyceride) and SiE (silyl ether) were observed only in BT. This information is very useful for the differentiation of lard, BT, and CF in order to avoid adulteration and mislabelling. The details of the specific lipid groups found in lard, CF, and BT with their lipid compounds for each group are presented in Table 1. Previous research on the discrimination of lard from other fats such as chicken fat, goat fat, and cattle fat has been performed based on fatty acid profiles using GC-TOF-MS. It was found that three fatty acid methyl esters of methyl trans-9,12,15-octadecatrienoate (C18:3 n3t), methyl 11,14,17-eicosatrienoate (C20:3 n3t) and methyl 11,14-eicosadienoate (C20:2 n6) could be used as potential discriminating lipids of lard from other animal fat samples (Indrasti et al., 2010). However, it is only capable of identifying fatty acids, not the comprehensive types of lipids. Another study aimed to apply different analytical approaches such as gas liquid chromatography (GLC), HPLC, and differential scanning calorimetry (DSC) to discriminate lard from beef tallow, mutton tallow, and chicken fat. The GLC method was not suitable for discriminating lard from the others by using overall fatty acid compositions. Triacylglycerol (TAG) análisis using HPLC showed a TAG profile for lard that differs from beef tallow and mutton tallow, but similar to chicken fat. The analysis of lard using DSC showed a different melting temperature for lard compared to other animal fats, although further analysis is still required in order to be more specific (Marikkar et al., 2021).

Overall, liquid chromatography-high resolution mass spectrometry using the Orbitrap mass analyzer could be used for the comprehensive identification of lipid compositions in lard, BT, and CF. Some differences in the lipid groups were detected, which is important for the differentiation of lard from BT and CF. The chemometric analysis could be used to identify the metabolite pattern, in this case lipids, to differentiate and classify lard, beef tallow, and chicken fat.

Lipidomic analysis using LC-HRMS could be used to detect the presence of lard adulteration both in BT and CF at a ratio of 50% adulteration. The TIC of adulterated CF and BT with 50% lard was still similar to samples of pure BT and pure CF (data not shown). The main composition of lipid groups in adulterated BT and CF with lard, such as triglycerides, followed by diglycerides and ceramides, was similar to pure samples. Investigations using lipid compositions showed that BT and CF adulterated with lard could be differentiated from pure BT and CF samples. The specific lipid groups in lard could be detected in adulterated imples of BT, namely LPC (lysophosphatidylcholine), MePC (monoetherglycerophosphocoline), PC (phosphatidylcholine), and PI (phosphatidylinositol). These lipid groups were absent from pure BT. Therefore, it can be used to indicate the presence of lard in BT. At the same time, in adulterated CF with 50% lard, the specific lipids of lard which were absent from CF such as biotinylPE (biotinylphosphoetanolamine), PC, LPC, MePC and DG were detected.

Figure 3A shows the Venn diagram of lipid metabolites between pure lard, pure BT and adulterated BT with 50% lard. The results showed that 99 lipid compounds were present in lard, BT and adulterated BT. 161, 119, and 85 lipid compounds were found specific to BT, lard, and adulterated BT, respectively. On the other hand, the results of the Venn diagram from pure lard, pure CF and adulterated CF with 50% lard as depicted in Figure 3B show that 98 lipid compounds were found only in lard, 188 lipids were specific to CF, and 78 lipids were observed only in adulterated CF. These lipids could be used to identify the authentication purposes of CF from lard. Meanwhile a number of 95 lipid compounds were found in lard, CF, and CF adulterated with lard.

The chemometric analysis using PCA successfully differentiated between pure samples of BT and CF and the adulterated ones using lard as shown in the PCA score plot in Figure 4A. The PCA performed with six principal components successfully differentiated adulterated samples from pure samples with R² = 0.999 and Q² = 0.996. A high R² value indicated high model accuracy, whereas a high value for Q² (> 0.500) showed good model predictability. (Bevilacqua et al., 2017). All adulterated samples of BT and CF with 50% lard appeared around the score plot for lard. PLS-DA using three components was successfully used for discrimination and classification between pure and adulterated samples of BT and CF with lard as depicted in the PLS-DA score plot in Figure 3B. The goodness of fit of the PLS-DA model was shown by R²X (0.706) and R²Y (0.988) values. Meanwhile, the Q² value (0.976) demonstrated the good predictability of the model. In addition, all the adulterated samples of BT and CF could be correctly classified as adulterated samples with 100% accuracy. The PLS-DA model was evaluated by means of a permutation test and ROC value to validate the PLS-DA model as shown in Figure 5. The permutation test used 999 permutations to confirm the validity of the PLS-DA model. All permutated models on the left side were lower than the original models on the right side (Figure 5A). In addition, the intercept of Q2 was zero and lower than zero (0.0, -0.43), thus indicating good model validity. The analysis of ROC was evaluated using the area under the curve (AUC) value. The resulting AUC value was 1 for each class (Figure 5B), which confirmed the validity of the model (Rivera-Pérez et al., 2021). In addition, the analysis of variable importance projections (VIP) value in PLS was used to identify potential lipids which play important roles in discriminating between pure fat samples (BT and CF) and those adulterated with lard. Table 2 shows the potential lipid biomarkers obtained from the VIP analysis. Variables with a VIP value greater than 1 are considered important variables as potential biomarkers for sample discrimination. Most of them were glycerolipids (DG and TG).

Previous research on lipidomic analysis using DART-TOF-MS (direct analysis in real time-time of flight-mass spectrometry) has been successfully used for the authentication of beef tallow. This research focused on triacylglycerol (TAG) compositions. A chemometric linear discriminant analysis (LDA) using TAG compositions was performed to successfully discriminate between pure and adulterated beef tallow samples with lard (Vaclavik et al., 2011). Our study provided more comprehensive lipid compounds because it is focused not only on the TAG compositions but a wider range of lipid compounds as well.