Simultaneous analysis of patin fish oil (Pangasius micronemus) and bandeng (Chanos chanos) fish oil using FTIR spectroscopy and chemometrics

Ikhsan, A.N., Irnawati, I., Lestari, L.A., Erwanto, Y. and Rohman, A. Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Gadjah Mada University, Yogyakarta, 55281 Indonesia Faculty of Pharmacy, Halu Oleo University, Kendari, 93232, Indonesia Center of Excellence, Institute for Halal Industry and System (IHIS), Gadjah Mada University, Yogyakarta, 55281, Indonesia Department of nutrition and health, Faculty of Medicine, Public Health and Nursing, Universitas Gadjah Mada, Yogyakarta 55281, Indonesia Division of Animal Products Technology, Faculty of Animal Science, Universitas Gadjah Mada, Yogyakarta 55281, Indonesia


Introduction
Patin fish (Pangasius micronemus) and bandeng fish (Chanos chanos) are cultivated in Indonesia. Besides their meat, fish oils also have a huge potency to be developed as a dietary supplement (Siscovick et al., 2017). Patin fish oil (PFO) and bandeng fish oil (BFO) were composed mainly of polyunsaturated fatty acids (PUFA). PFO contained a high level of omega-6 fatty acid while BFO contains a high level of omega-3 fatty acid. In the future, PFO and BFO may be combined as food supplements. Fish oil supplementation provides some beneficial effects on human health such as preventing cardiovascular disease and is also essential for child brain development (Ramaswami, 2016;Gao et al., 2017;Hansen et al., 2017).
Verifying the quality and identity of each fish oil was needed for the following reasons: well-defined quality, legal compliance, dosage, and product safety (Danezis et al., 2016;Bansal et al., 2017). Conventional methods were used to establish reference materials to analyse adulteration and guarantee quality (Danezis et al., 2016;Sabir, et al., 2017;He et al., 2020;Gao et al., 2021). However, the fish oil components were composed of some complex components whose reference materials were hard to find. Multivariate analysis based on specific fish oil characteristics was a more conclusive method to discriminate each component (Jha et al., 2016;Avramidou et al., 2018;Mohammed et al., 2021). Authentication of different oils and fats has been performed using multivariate analysis between triacylglycerol and fatty acid fingerprints using gas chromatography (GC) and proton transfer reaction mass spectra (PTR-MS) (van Ruth et al., 2010).
FTIR spectroscopy was a common method to analyse fish oil characteristics based on their fingerprint in nature (Poonia et al., 2017;Valand et al., 2020). This method was reported to be effective for distinguishing each oil by its spectral difference (Rohman, 2017). Combined with partial least square (PLS) and principal component regression (PCR), the FTIR spectroscopy method has been claimed effective to authenticate patin fish oil with palm oil as an adulterant (Putri et al., 2019), cod liver oils from other fats and oils as adulterants (Rohman and Che Man, 2009). Besides that, it was also claimed effective to authenticate virgin oils with paraffin oils as an adulterant (Amit et al., 2020) and avocado oil in ternary mixtures with sunflower and soybean oils (Jiménez-Sotelo et al., 2016). However, the study related to the simultaneous analysis of 2 different fish oils in a mixture has not been conducted. The novelty of this research was to perform simultaneous analysis between 2 different fish oil in a mixture, which was never conducted before.
Furthermore in this research, FTIR spectroscopy combined with PLS and PCR methods was used to conduct a simultaneous quantitative analysis of PFO and BFO. Based on the literature review process, there were no sufficient findings related to combining FTIR spectroscopy with PLS and PCR for quantitative analysis of PFO and BFO simultaneously. Hence, the main objective of this research was to quantitatively analyze and optimize PFO and BFO using FTIR spectroscopy combined with PLS and PCR simultaneously.

Materials and methods
Bandeng fish oils and patin fish oils were obtained from a local fish breeder in Pati, Central Java, Indonesia. Hexane and acetone pro-analytical grades for sampling cleaner were purchased from E. Merck (Darmstadt, Germany).

Sample preparation
Both fish were descaled, filleted, and cleaned. Three kilograms of clean fish flesh were placed in an aluminium tray. Samples were dried using a cabinet dryer at 50°C for 24 hrs. The extraction method was adapted from the previously reported method by Honold et al. (2016). Dried samples were placed into a pressing chamber covered with a filter cloth. The pressing process was performed using 100 kN force for 2 mins. The extracted oil was centrifugated at 5000×g for 10 mins to separate the sediment.

Preparation of calibration and validation samples
PFO and BFO were mixed into binary mixtures with accumulated levels of 100%. The total volume of binary mixtures was 2 mL, while the total amount of samples was 25. For instance, sample 1 was composed of 96.5% of PFO and 3.5% BFO. The validation samples were prepared independently using the same spanned concentration from 0-100 % as in calibration samples. The selection of its percentage in Table 1 was based on random order as suggested by Excel software (Microsoft Inc., USA).

FTIR spectroscopy analysis
All samples were scanned using an FTIR spectrophotometer (Thermo Scientific Nicolet iS10, Madison, WI) according to Irnawati et al. (2020). Meanwhile, the obtained spectra were processed using Omnic software. The samples were measured in 11 different wavenumber regions between 4000-650 cm -1 with a resolution of 16 cm -1 , and scanning numbers of 25 replicates (2 for calibration and 1 for validation). Horizontal Attenuated Total Reflectance (HATR) composed of ZnSe crystal was used as used sampling accessory. A correcting action was carried out by scanning a new reference air as background after every sample scanning. Triplicate data points were used to make a correlation between the predicted value and FTIR spectra. The spectra data were recorded as absorbance values. software was used for chemometrics analysis. Partial least square (PLS) and principal component regression (PCR) were used to make a correlation between actual values of PFO and BFO with predicted values and using FTIR spectra. For quantitative analysis, 25 mixtures of samples containing PFO and BFO were analysed using spectral regions between 4000 cm -1 to 650 cm -1 . The normal, 1 st derivative and 2 nd derivative spectra were observed to increase spectral resolution. The root mean square error of calibration (RMSEC), root mean square error of cross prediction (RMSEP), coefficient of determination for calibration (R cal 2 ), and coefficient of determination for validation (R val 2 ) were calculated using, TQ analyst software.

Results and discussion
Patin fish oil (PFO) and bandeng fish oil (BFO) were chosen based on their popularity in Indonesia and their nutritional values (Sugata et al., 2019;Ilza and Sukmiwati, 2020). PFO contains high omega-3 while BFO contains high omega-6. The mixture between PFO and BFO was predicted to be developed as a dietary supplement in the future. Sample binary mixtures were scanned in wavenumber regions between 4000 cm -1 and 650 cm -1 . Figure 1  Optimization of partial least square regression (PLS) and principal component regression (PCR) can be observed in Table 2 and Table 3. The selected method for (PFO) simultaneous analysis was based on the highest values of coefficient of determination for calibration (R cal 2 ) and coefficient of determination for validation (R val 2 ), followed by the lowest values of root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) in Table 2. The first derivative FTIR spectra at wavenumbers of 3050-2800 and 1800-1700 and 1500-650 cm -1 were the highest R cal 2 values (0.9991); R val 2 (0.9990); followed by the lowest RMSEC (0.0135); and the lowest RMSEP (0.0146) using PCR. Moreover, using PLS as the selection method resulted in the first derivative FTIR spectra at wavenumbers of 3050-2800 and 1800-1700 and 1500-650. The R cal 2 ; R val 2 ; RMSEC; and RMSEP value was 0.9998; 0.9994; 0.0072; and 0.0121, respectively. Meanwhile, the selected method for (BFO) simultaneous analysis was based on the highest values of coefficient of determination for calibration (R cal 2 ) and coefficient of determination for validation (R val 2 ), followed by the lowest values of root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) in Table 3. Similar to PFO above, the first derivative FTIR spectra at wavenumbers of 3050 -2800 and 1800-1700 and 1500-650 cm -1 were the highest R cal 2 values (0.9991); R val 2 (0.9990); followed by the lowest RMSEC (0.0135); and the lowest RMSEP (0.0146) using PCR. Furthermore, using PLS as the selection method resulted in the first derivative FTIR spectra at wavenumbers of 3050-2800 and 1800-1700 and 1500-650. The R cal 2 ; R val 2 ; RMSEC; and RMSEP value was 0.9998; 0.9994; 0.0072; and 0.0121, respectively. Putri et al. (2019) conducted an authentication of patin fish oil from palm oil using FTIR spectroscopy combined with chemometrics. The method provided perfect discrimination (100%) of patin fish oil from palm oil with low errors (RMSEC = 0.805 and RMSEP = 2.22) and high accuracy (R 2 > 0.999). This research has confirmed that FTIR spectroscopy combined with chemometrics (PLS and PCR) can be used as a (2019) research were slightly higher than in this research. It may be caused by the differences in fatty acid composition between patin fish oil and palm oil. The patin fish oil was composed mainly of unsaturated fatty acid, however, palm oil is mainly composed of saturated fatty acid (Putri et al., 2019). Compared with this research, both PFO and BFO are mainly composed of unsaturated fatty acids, so PFO and BFO have similar spectra to patin fish oil and palm oil. Figure 2 presents the actual and calculated values correlation of PFO and BFO binary mixtures. Based on the chart, the errors that occurred in modelling were distributed randomly along the correlation line. Moreover, the errors obtained were not systematic. The highest values R cal 2 and R val 2 , but the lowest values of RMSEC and RMSEP indicate the selected method performed accurate results for the analysis of PFO and BFO binary mixture simultaneously (Miller and Miller, 2018

Conclusion
The combination of Fourier transform infrared spectroscopy (FTIR) with partial least square (PLS) and principal component regression (PCR) had successfully been used for simultaneous analysis of binary mixtures of patin fish oil (PFO) and bandeng fish oil (BFO). The simultaneous analysis method was based on variable absorbances values of the first derivative spectra at 3050-2800, 1800-1700, and 1500-650 cm -1 for PFO and BFO.

Conflict of interest
The authors declare no conflict of interest. Gao, B., Xu, S., Han, L. and Liu, X. (2021 Table 3. Optimization of principal component regression (PCR) dan partial least regression (PLS) methods to determine (BFO) component in different wavelength areas for simultaneous analysis.