Determination of acid, peroxide, and saponification value in patin fish oil by FTIR spectroscopy combined with chemometrics

Putri, A.R., Rohman, A., Setyaningsih, W. and Riyanto, S. Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Universitas Gadjah Mada, Yogyakarta, Indonesia 2 Institute of Halal Industry and Systems, Universitas Gadjah Mada, Yogyakarta 55281 Indonesia Department of Food and Agricultural Product Technology, Faculty of Agricultural Technology, Universitas Gadjah Mada, Jalan Flora No. 1, Bulaksumur, Depok, Sleman 55281 Yogyakarta, Indonesia


Introduction
Patin (Pangasius micronemus) is a freshwater fish in Indonesia and have exported for years to the western market. Indonesia is the second largest of patin producer, with production reaching up to 16.1% of the total production of patin in the world (Ramadhan et al., 2016). Hashim et al. (2015) have sucessfully extracted fatty acid from patin and get lipid up to 31% with the level of omega-3 was 4.7%. Hence, patin can be a good source of omega-3. While in Thailand and Vietnam, patin was extracted to obtain the fish oils and the patin fish oils (PFO) have been widely traded.
In the market, that is important to know the quality of fish oil. Some parameters used to determine the quality of fish oil. To determine the quality of fish oil can be performed by determining of physico-chemical values such as acid, peroxide, and saponification value (Rohman et al., 2015). Titration method was used to determine the acid, peroxide, and saponification value, based on the chemical reaction between free fatty acids and the reagent. But, the titration method has some disadvantages such as being time-consuming and laborintensive, requiring large amounts of organic solvents, use of the highly toxic, and carcinogenic reagent, difficulty in distinguishing the end-point with samples containing coloured substances and largely dependent on the skills of the analyst (Rohman et al., 2012;Jiang et al., 2016).
Currently, Fourier transform infrared spectroscopy (FTIR) combined with chemometrics has been widely used for determining acid, peroxide, and saponification value. FTIR has been used for determination of acid value on edible oil-based on -O-H stretching (Jiang et al., 2016), peroxide value on edible oils based on -COOstretching (Yu et al., 2009;Hu et al., 2019), to monitoring peroxide value in oxidized emulsions (Hayati et al., 2005), and saponification value on red fruit oil using partial least square calibration (Rohman et al., 2015).
However, the application of FTIR spectroscopy for the determination of acid, peroxide, and saponification value of patin fish oil (PFO) has not been reported yet. Therefore, FTIR spectroscopy combined with principal component regression (PCR) and partial least square (PLS) was developed for the determination of acid, peroxide, and saponification value of PFO from some extraction methods.

Extraction method
Extractions of PFO were performed using maceration method with wet and dried samples and soxhlet method also with three variations of solvent are n -hexane, ethyl acetate, and chloroform. The patin flesh was cutting into small pieces then extracted with the solvent for maceration with wet samples. Patin flesh was dried in the oven at 60°C for 24 hrs before extracted using maceration with dried samples and Soxhlet method. Maceration process was performed at room temperature while the Soxhlet method at 82°C. After the extraction step, the solvent was evaporated using the rotary evaporator at 50°C to get the PFO.

Acid value determination
Determination of acid value using the AOCS official method Cd 3a-63 (AOCS, 2004). The oil sample (2 g) was mixed with 20 mL ethanol and added with 2 mL phenolphthalein in a 250 mL Erlenmeyer flask. The mixture was titrated with 0.01 M of KOH and shaken vigorously until the color was changed (from white to pink). The result was expressed as the number of milligrams of KOH required to neutralize the free fatty acid in 1 g of the sample.

Peroxide value determination
Peroxide value is a measure of peroxides contained in the oil and is determined by measuring iodine released from potassium iodide. Determination of peroxide value was performed according to AOCS official method Cd 8b-90 (AOCS, 2005). The oil sample (5±0.01 g) was dissolved in 30 mL acetic acid-chloroform (3:2) solution. Then saturated KI solution and distilled water were added and shake the flask vigorously to liberate iodine from chloroform layer. The mixture was titrated with 0.01 N sodium thiosulphate using starch solution as an indicator.

Saponification value determination
The saponification value was expressed as the number of milligrams of potassium hydroxide (KOH) required to saponify 1 g of oil. Determination of saponification value was according to AOCS method Cd 3-25 (AOCS, 1990). The 2 g of oil were dissolved with ethanol in Erlenmeyer flask. Then connected with an air condenser and boil gently for 1 hr in order that sample is completely saponified. After it cooled and added 1 mL phenolphthalein. The mixture was titrated with 0.5 N HCl until the pink colour has just disappeared.

FTIR analysis
Samples were analysed using FTIR spectrometer (Thermo Scientific Nicolet iS10, Madison, WI) using Omnic software. The measurements were done in the middle infrared region of 650-4000 cm -1 with 32 scans and at the resolution was 16 cm -1 . The background scan was performed to reduce the effect of the reference spectrum of the air. Before and after analysis sample, ATR crystal was cleaned with acetone p.a. Replications were done with scanning the same samples for three times.

Data analysis
Multivariate analysis of the acquired data was performed using principal component regression (PCR) and partial least square (PLS) with TQ Analyst software version 9 (Thermo Fisher Scientific Inc.) while the statistic analysis was done using Minitab 18. In this research, Principal component regression (PCR) and partial least square (PLS) were used to determine the physico-chemical value from FTIR spectral. PCR is one of inverse calibration technique in which the physicochemical value from titration method (x-axis) is used as a predictor, while responses such as absorbance at several wavelengths are located on the y-axis. PCR performs multiple inverse calibrations of predictor variables against the scores (knowns as principal components) rather than original variables . Whereas the PLS method uses a correlation between changes in spectral absorption and sample concentration then computing with other spectra that can disturb the analyte spectra (Ballabio and Todeschini, 2009). In this research using normal, first derivative and second derivative spectra to identify the targeted spectra from the compounds by increasing the spectral resolution .
The collected data set was divided into calibration and validation subset. Prediction models were validated with cross validation 'leave one out' technique. Cross validation evaluates the data by excluding selected samples in the regression model and then constructing a new model for the remaining samples. The new model was evaluated and the error values for predicted observations are computed. The new samples are then FULL PAPER excluded from the model set and a new model is constructed. This procedure is repeated until all samples in the PCR and PLS models have been excluded once . In order to compare the performances of the developed calibration models and express their predictive ability, two statistical values, namely root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) were calculated. Among these values, RMSEP gives the best estimation of future performance of the calibration model (Temiz et al., 2017).

FTIR Spectra of PFO
The FTIR spectra of PFO was shown in Figure 1. Each peak corresponded to a functional group of fatty acid structure. The peak at wavenumber 721 cm -1 was due to the rocking vibration of methylene (-CH 2 ), while the peaks at 1114 and 1234 cm -1 were from -C-O vibrations. The bending vibration of methylene was observed at wavenumber 1461 cm -1 . Meanwhile the carbonyl (C=O) stretching vibration was observed at a wavenumber 1744 cm -1 . The peaks at wavenumbers 2854 and 2921 cm -1 represent asymmetric and symmetric stretching vibration of methylene (-CH 2 ) respectively (Rohman et al., 2011;Putri et al., 2019).

Determination of AV
The acid value (AV) is an explanatory parameter for evaluation of the level of hydrolysis of oil. Free fatty acid and glycerol were the results of the hydrolysis reaction. The acid value usually can be determined by acid/base titration (Mahboubifar et al., 2016). The AV from both titration and FTIR methods were shown in Table 1. The AVs from FTIR were obtained from optimization method (Table 2). Based on the ANOVA test using Minitab 18, p-value from both methods was 0.00. So, the result from both methods not significantly different (p<0.05). The AV from the Soxhlet method were higher than in the maceration method. According to FAO about the standard for fish oils (2017), the acceptable of AV on fish oil is ≤ 3 mg KOH/g. The high of AV can be caused by heat treatment while in the extraction process. The heat treatment cause oxidation and hydrolysis on PFO to produce free fatty acid (Herchi et al., 2016).
According to Table 2, the PCR technique gave the best result than others. The wavenumber of 2000-3100 cm -1 was chosen to determine AV because give the highest coefficient determination for calibration and validation, 0.9879 and 1.000 respectively with the lowest RMSEC and RMSEP were 0.210 and 0.972. Figure 2 presents the linear regression of the PCR method expressing the relationship between AVs from the titration method (as actual value) and FTIR spectral (as calculated value). The peak at wavenumber of 2000-3100 cm -1 corresponding to the stretching vibration of cis -C=H at 3006 cm -1 and vibration of CH 2 stretching at the peak at 2854 and 2921 cm -1 from free fatty acid (Che   (Figure 3).

Determination of PV
Peroxide value (PV) is used to measure the level of peroxide/hydroperoxide formed at the initial process of oil oxidation by measuring iodine released from potassium iodide on titration process (Nunes, 2014;Naz and Saeed, 2018). The PVs of PFO by titration and FTIR combined with PCR method was shown in Table 1. Based on the ANOVA test, the PVs from titration and FTIR method were not significantly different (p<0.05). The accuracy of iodometric titration to determine the PVs depends on some other experimental factors such as precise timing and protection of the reaction mixture from oxygen (Armenta et al., 2007)  the mistake of distinguishing on the endpoint (Jiang et al., 2016). Peroxide/hydroperoxide which is intermediate species. These intermediates are unstable species that can react very quickly that make the mistake when the determination of endpoint (Mahboubifar et al., 2016). The acceptable of PV for fish oil is ≤ 5 meq O 2 /kg oil (FAO, 2017). The oxidation rate increase with the temperature results high amount of peroxide, that make PVs of PFO from the Soxhlet method were highest than others (Popa et al., 2017).
Based on the optimization method, PLS with first derivative spectra showed the best result in the prediction of the desired response (Table 3). PLS at wavenumbers of 1500-2111 cm -1 were selected for determining the PVs because it gives the highest coefficient value for determination and validation. The coefficient value (R 2 ) of determination for the calibration and validation were obtained at 0.9982. The RMSEC and RMSEP value were 0.200 and 0.564 respectively. The excellent linear relationship between the PVs measured by titration and FTIR method shown in Figure 4. Figure 5 presents the area 1500 -2111 cm -1 is attributed for -C=O ester stretching at wavenumber of 1700 cm -1 (Che . According to Guillén and Cabo (1999), under the oxidative conditions, the oils have a very intense absorption due to the ester carbonyl functional group of the triglycerides causes a peak at area 1746 cm -1 . They also found that secondary oxidation products that cause an absorption at 1728 cm -1 which overlaps with the peak of the ester functional group.

Determination of SV
Saponification value (SV) is an index of the average molecular mass of fatty acid in the oil sample (Zahir et al., 2017). The SV of PFO was determined by titration and FTIR combined with PCR and PLS methods were shown in Table 1. Both methods give not significantly different result (p<0,05). From the optimization method, the area at wavenumber of 1007-1600 cm -1 was chosen to determine PVs   Rohman et al. (2015) that use absorbance at wavelength of 1400-1600 cm -1 with PLS technique to determine the SVs. The linear regression as shown in Figure 6. First derivative spectra were displayed in Figure 7.

Conclusion
In this work, the AV, PV, and SV of PFO were successfully determined via FTIR spectroscopy combined with chemometrics using PCR and PLS method. The rapid determination of AV, PV, and SV by FTIR spectroscopy is, therefore, suitable and practical option for process control. Another advantage of FTIR spectroscopy method is that it is environmentally friendly as no chemical is needed except acetone for cleaning the ATR crystal. By utilizing this method, the chemical cost is negligible as compared to the AOCS standard method.
Determination of edible oil parameters by near infrared spectrometry. Analytica Chimica Acta, 596