- Email: [email protected]
Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed
FRIN-05370; No of Pages 7 Food Research International xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Food Research International journal homepage: www.elsevier.com/locate/foodres
Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed J. Azmir a, I.S.M. Zaidul a,⁎, K.M. Sharif a, M.S. Uddin a, M.H.A. Jahurul b, S. Jinap c, Parvaneh Hajeb c, A. Mohamed d a
Faculty of Pharmacy, International Islamic University Malaysia, Kuantan Campus, 25200 Kuantan, Pahang, Malaysia Department of Food Science and Nutrition, Faculty of Applied Sciences, UCSI University, 56000 Kuala Lumpur, Malaysia Food Safety Research Centre (FOSREC), Faculty of Food Science and Technology, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia d Faculty of Pharmacy, Cyberjaya University College of Medical Sciences, 63000 Cyberjaya, Malaysia b c
a r t i c l e
i n f o
Article history: Received 24 April 2014 Received in revised form 22 June 2014 Accepted 28 June 2014 Available online xxxx Keywords: Phaleria macrocarpa seed oil Fatty acids Response surface methodology Artificial neural network
a b s t r a c t Good quality oil with high unsaturated fatty acids was found in the seed of a medicinal plant Phaleria macrocarpa (Mahkota dewa). Different parts especially fruit flesh of this plant are being traditionally used as important folk medicine whereas seed of this plant is usually neglected. In this study, the oil was extracted from P. macrocarpa seed using supercritical carbon dioxide. The extraction parameters were optimized by central composite design (CCD) of response surface methodology (RSM). Due to the non-linearity of the extraction process, artificial neural network (ANN) was also applied for predicting the oil yield. The optimum conditions obtained from RSM were 72 °C, 42 MPa and 4.5 ml/min CO2 flow rate where the oil yield was 52.9 g per 100 g of dry sample and coefficient of determination (R2) was 0.99. The ANN and RSM prediction showed similar R2 of 0.99 and ANN has lower average absolute deviation (AAD) of 0.25% compared to RSM (AAD of 0.31%). Five fatty acids were identified by gas chromatography–mass spectroscopy (GC–MS) analysis of the oil. The amount of oleic acid (18:1) was found to be highest (43.56%) among all the fatty acids. The total unsaturated fatty acid was 73.62% and saturated fatty acid was 26.38% in the P. macrocarpa seed oil. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Phaleria macrocarpa (local name: Mahkota dewa), member of Thymelaeaceae family has been traditionally treated as an important medicinal plant for centuries in Malaysia and Indonesia. Mahkota dewa means “God's Crown” meaning it descends from heaven to help mankind. Leaves, stem, fruit and seed of this plant have been examined extensively for their biological and pharmacological potencies (Ali et al., 2012; Hendra, Ahmad, Oskoueian, Sukari, & Shukor, 2011; Susilawati, Matsjeh, Pranowo, & Anwar, 2011; Zhang, Xu, & Liu, 2006). Apart from bioactive compounds responsible for pharmacological activities, the seed of this plant contains high amount of valuable oil reported by Azmir et al. (2014). Solvent extraction has been commonly used for the extraction of oil from plant matrix. It has, however, some inherent drawbacks; high heat is essential for the distillation of the oil, the oil contains residual solvents, and rancidity occurs during the separation process because the oil is oxidatively unstable (Scalia, Giuffreda, & Pallado, 1999). For the extraction of oil, supercritical carbon dioxide (scCO2) attracts considerable attention as a promising alternative to conventional organic ⁎ Corresponding author. Tel.: +60 9 570 4841; fax: +60 9 571 6775. E-mail address: [email protected] (I.S.M. Zaidul).
solvent extraction. Supercritical carbon dioxide has high extraction efficiency of oil and it is non-toxic, non-flammable, environmentally friendly, non-explosive, cost effective, less laborious, available and easy to remove from the extract thus eliminates post processing steps (Azmir et al., 2013; Lang & Wai, 2001). The extraction efficiency of scCO2 can be influenced by many factors such as extraction temperature and pressure, particle size and moisture content of feed material, time of extraction and flow rate of CO2 (Ibañez, Herrero, Mendiola, & Castro-Puyana, 2012; Temelli & Güçlü-Üstündağ, 2005). Response surface methodology (RSM) is a very effective technique to optimize significant operating parameters as well as to determine the effect of individual input variable and the interaction of variables for maximizing the benefits (Raymond & Montgomery, 2002; Wang, Liu, Wei, & Yan, 2012). Optimum and valid output can be produced with minimum effort, less time and resources by RSM as it requires minimum experimental runs for optimizing multiple independent variables (Sharif et al., 2014). A polynomial equation is used to describe and predict the response variable in RSM. On the other hand, artificial neural network (ANN) is very helpful to predict the response for non-linear system like extraction. ANN is a simplified model of the structure of a biological network (Mandal, Sivaprasad, Venugopal, & Murthy, 2009) and an artificial neuron is its main processing element. The neuron receives input data from various sources, combines the
http://dx.doi.org/10.1016/j.foodres.2014.06.049 0963-9969/© 2014 Elsevier Ltd. All rights reserved.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
Pressure Controller
Valve
Cooler
Temperature Controller
Back Pressure Regulator
CO2 Pump
Extraction Vessel
data, generally performs a nonlinear operation on the result and then gives output as the final result (Baş & Boyaci, 2007; Ibrić, Djuriš, Parojčić, & Djurić, 2012). The main advantage of ANN is that it does not need any mathematical modeling (Mandal et al., 2009) like linear and polynomial regressions based modeling. The aim of this study was to extract the P. macrocarpa seed oil using scCO2. RSM and ANN were employed to model the extraction process and to compare the prediction quality. RSM with central composite design (CCD) was employed to optimize the process parameters including extraction temperature, pressure and CO2 flow rate. The analysis of fatty acids in terms of triglycerides composition of the extracted P. macrocarpa seed oil was also aimed in this study.
Valve
2
2. Method and materials Collector
2.1. Materials
(a)
CO2 Cylinder
The fresh fruits of P. macrocarpa were collected from a local supplier in Kuantan, Malaysia. The sample specimen was deposited in herbarium of International Islamic University Malaysia (IIUM), Kuantan, Malaysia. All reagents (analytical and HPLC grade) were purchased from Merck Ltd. (Darmstadt, Germany). The CO2 gas was purchased from Fuelink Marketing Sdn. Bhd. (Selangor, Malaysia).
Hidden layer
2.2. Extraction of oil by conventional method
Temperature Oil from P. macrocarpa seed was extracted by solvent extraction technique using n-hexane at optimum conditions of 72 °C for 8.4 h and solvent-to-feed ratio of 10.9 ml/g as reported in our previous study (Azmir et al., 2014). The total oil yields were determined in percentage of 100 g of ground seed sample on dry weight basis.
CO2 flow rate
2.3. Supercritical carbon dioxide (scCO2) extraction The schematic diagram of scCO2 apparatus used in this study is shown in Fig. 1a. The apparatus has a CO2 pump (model PU-1580 and model PU-2080, Jasco Corporation, Tokyo, Japan) and the pump was fitted with a cooling jacket to cool the CO2 supplied from the tank. Ethylene glycol–deionized water mixture (50:50, v/v) was circulated through the cooling jacket using a low-temperature bath circulator (model 631D, Tech-Lab manufacturing Sdn. Bhd., Selangor, Malaysia) to chill coolant down to −6.5 °C. For every experiments, 10 g of sample was loaded into a 50 ml extraction vessel (model EV-3, Jasco Corporation, Tokyo), and the vessel was maintained at the desired temperature. A back-pressure regulator (BPR) (model BP-1580-81, Jasco Corporation, Tokyo, Japan) was used to control the extraction pressure. The seed oil was separated from the supercritical fluid by pressure reduction through an expansion valve. A 20 ml blue cap schott bottle was used for collecting the oil. The extraction process was continued for 3 h. The total yields were determined as percentage on dry weight basis of sample as described in the Eq. (1). Yield ð%Þ ¼
Massextracted oil 100: Massseed powder
Yield
Pressure
ð1Þ
2.4. Response surface methodology (RSM) RSM with CCD was employed to determine the best combinations of variables for obtaining high extraction yield in scCO2 process. Three parameters, temperature (X1), pressure (X2), and flow rate of CO2 (X3) were considered as independent variables. The selection of variables and their level (temperature, 60 to 80 °C; pressure, 25 to 45 MPa and CO2 flow rate, 3 to 5 ml/min) was based on the literature review, and trial-error (data not shown). The formulated experimental design was required 20 experimental runs for this experiment. For estimation of
(b) Fig. 1. (a) Schematic diagram of the SFE apparatus and (b) Schematic structure of studied ANN model.
“pure error” six replicates were performed at the central point (0,0,0) of the design. The run order and the combination of variables are shown in Table 1. A second-order polynomial regression equation was used for predicting the response variable (Y) as Eq. (2). Y ¼ β0 þ
X
βi X i þ
X
2
βii X i þ
XX
βij X i X j
ð2Þ
Where, Y is the response variable, β0 is a constant, and βi, βii and βij represent the linear, quadratic and interactive coefficients, respectively. Xi and Xj are the independent variables. The Minitab software (version 16) was used for multiple regression analysis, analysis of variance (ANOVA), and determination of coefficient R2 which measures the goodness of fit of regression model. The t-value was also included for the estimated coefficients and associated probabilities. The total error criteria with a confidence level of 95.0% were the basis for test of statistical significance. 2.5. Artificial neural network (ANN) The 20 experimental data used for RSM were also used for ANN to predict the response output of oil yield. The total experimental data was randomly split into two data sets: 17 for training and 3 for validation. The training data sets were applied for computing the network
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
3
Table 1 The experimental and predicted oil yield obtained from RSM and ANN. Temperature (°C)
Pressure (MPa)
CO2 flow rate (ml/min)
Experimental yield (g) (Mean ± sd)
Predicted yield (g) (RSM)
Predicted yield (g) (ANN)
80 70 80 70 60 80 70 60 60 60 70 80 70 70 70 60 70 70 70 80
25 35 25 35 45 45 35 45 25 25 35 45 35 35 45 35 25 35 35 35
3 4 5 4 3 5 4 5 5 3 4 3 5 4 4 4 4 4 3 4
26.35 47.08 28.91 47.56 50.00 53.05 47.49 52.88 28.14 24.61 47.23 50.75 48.86 47.18 51.87 46.90 27.67 47.33 45.35 47.76
26.28 47.32 29.08 47.32 49.84 53.03 47.32 52.96 28.21 24.64 47.32 50.69 48.54 47.35 52.02 46.86 27.44 47.35 45.59 47.72
26.32 47.34 28.91 47.34 50.07 53.14 47.34 52.90 28.21 24.70 47.34 50.77 48.84 47.34 51.66 46.73 27.53 47.34 45.59 47.66
parameters and the validation data sets were used for ensuring the robustness testing of the network parameters. A feed forward ANN which uses back-propagation algorithm, based on a multi-layer perceptron (MLP) was applied using Peltarian Synapse (Version 1.6.0, the Peltarion Corporation, Stockholm, Sweden) for the modeling. MLP with three layers, the most popular ANN (Hussain, Bedi, & Singh, 1992; Jorjani, Chehreh Chelgani, & Mesroghli, 2008) which was consisted of an input layer, hidden layers and an output layer was used for this experiment. Extraction temperature, pressure and CO2 flow rate were the input factors and yield of oil was the output. The structure of ANN is shown in Fig. 1b. The transfer function (weight layer) that expresses an internal activation level is the constituent of neuron (Ahnert, Travençolo, & da Fontoura Costa, 2009). Sigmoidal function was used as transfer function to determine the output from a neuron by transforming its input. The input and output data were normalized to the range between 0 and 1 to avoid large magnitude of the data using Eq. (3) (Erzin, Rao, & Singh, 2008). Xnorm ¼
X−Xmin Xmax −Xmin
ð3Þ
Where, Xnorm is the normalized value, X is the actual value, Xmax is the maximum value and Xmin is the minimum value. The mean squared error (MSE) and R2 were used to test the performance of the developed network by Eqs. (4) and (5), respectively. MSE ¼
1 Xn 2 ðYi −Ydi Þ i¼1 n Xn
ð4Þ
2
ðYi −Ydi Þ R ¼ 1−Xni¼1 2 ðYdi −Ym Þ i¼1 2
ð5Þ
Where, n is the number of points, Yi is the predicted value obtained from the neural network model, Ydi is the actual value, and Ym is the average of the actual values. The prediction capabilities of the two techniques, RSM and ANN were compared by the R2 value and absolute average deviation (AAD). The AAD was calculated using Eq. (6) (Baş & Boyaci, 2007): AAD ¼
nhX
p i¼1
i o P 100: Yi; exp −Yi;cal Yi; exp
=
=
ð6Þ
± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±
0.04 0.08 0.03 0.20 0.23 0.12 0.15 0.08 0.06 0.06 0.03 0.07 0.06 0.08 0.02 0.20 0.07 0.15 0.10 0.16
Where, Yi,exp and Yi,cal are the experimental and calculated responses, respectively. P is the number of the experimental run. 2.6. Fatty acid composition analysis by GC–MS Fatty acids were derivatized properly before the GC–MS analysis and the derivatization was done according to the method of Ichihara and Fukubayashi (2010) with some modifications. The oil sample was taken for experiment after being melted at 60–70 °C and homogenized thoroughly. Commercial concentrated HCl (35% w/w) was diluted with 41.5 ml methanol to make 50 ml 8.0% (w/v) HCl reagent. The HCl reagent was stored in a refrigerator until use. A black screw cap glass test tube was used to take exactly 100 μl of oil sample and the sample was dissolved in 0.2 ml toluene. After that, 1.5 ml methanol and 0.3 ml of HCl reagent were added to the oil solution. The tube was vortexed and then heated at 95 °C for 90 min for rapid methylation. After cooling the oil solution at room temperature, 1 ml hexane and 1 ml water were added for the extraction of fatty acid methyl esters (FAMEs). The tube was vortexed again and the hexane layer with FAMEs was separated. For the composition analysis of FAMEs, an Agilent 6890 N gas chromatography coupled with Agilent MS-5973 mass selective detector (Agilent Technologies; USA, serial no. US14113031) was used. The prepared FAMEs were injected onto the HP-5MS column (30 m dimension, 0.25 mm i.d, 0.25 μm film thickness) using an Agilent autosampler 7683 series injector and the split ratio was 50:1. The initial oven temperature was maintained at 150 ºC for 2 min and increased at a rate of 4 °C/min up to 230 °C, then kept at 230 °C for 5 min. The injector and detector temperatures were held at 240 °C and 260 °C, respectively. Helium was the carrier gas used at a flow rate of 0.8 ml/min. A 70 eV electron impact (EI) in a 50–550 m/z scan range was applied for the operation of mass spectrometer (Kandhro et al., 2008). For the analysis, ChemStation integrated software (Agilent Technologies) was used. The fatty acids were tentatively identified by matching the mass spectra with NIST mass spectral database (NIST08). 3. Result and discussion 3.1. Analysis of RSM model The experimental oil yield for different combinations of input variables obtained from CCD is presented in Table 1. The input and response variables were analyzed to obtain the predicted response for
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
4
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
oil yield by the second-order polynomial equation as presented in Eq. (7).
(a)
2
Y ¼ 0:280 X 1 þ 6:750 X 2 þ 5:546 X 3 −0:001 X 1 −0:076 2 2 ð7Þ X 2 −0:289 X 3 −0:002 X 1 X 2 −0:019 X 1 X 3 −0:011 X 2 X 3 −117:636 The multiple regression coefficients of the model calculated by the least square technique are demonstrated in Table 2. The R2 of the model was 0.99 that indicates a good agreement between the experimental and predicted yield of oil. The lack of fit of this model was non-significant with the p value of 0.329 meaning that the second order polynomial equation was adequate for predicting the oil yield under all combinations of values of the factors. The linear and square effects of pressure were highly significant (p ≤ 0.01) where in both cases the effects of temperature were non-significant (p ≥ 0.05). The linear effect of CO2 flow rate was significant (p = 0.002) but the quadric effect was not significant (p = 0.073). The interaction between pressure and temperature and temperature and CO2 flow rate was significant (p ≤ 0.05) within the experimental ranges. The interactions of the studied factors on the oil yield are represented by three-dimensional response surface shown in Fig. 2. The interaction effect of temperature and pressure on the oil yield at a fixed CO2 flow rate of 4 ml/min is shown as surface plot in Fig. 2(a). At constant temperature, the yield increased significantly with pressure. Pressure showed positive linear effect at both low and high temperatures. On the other hand, temperature did not change the yield much neither at low pressure nor at high pressure. Density of scCO2 increases with pressure which improves the solubility and mass transfer of oil (Abbasi, Rezaei, & Rashidi, 2008). Fig. 2(b) represents the response surface and contour plot depicting the influence of temperature and CO2 flow rate on the oil yield at constant pressure (35 MPa). The flow rate of CO2 had a positive effect on the oil yield irrespective of the level of temperature whereas the temperature had positive effect only at low CO2 flow rate. At low flow rate, the increase of the temperature increased the yield of oil. When the flow rate reached to 5 ml/min, the effect of flow rate became more prominent on the oil yield that might suppressed the effect of the temperature. In contrast, the oil yield increased with CO2 flow rate in both, at high and at low temperatures. The possible explanation of this phenomenon is that the equilibrium state between solute and solvent could be favorable for the high mass transfer rate (Elkanzi & Singh, 2001; Kumoro & Hasan, 2007). At low CO2 flow rate, mass transfer resistance reduces the amount of oil transported to the CO2 where with the increase of CO2 flow rate the mass transfer resistance declines until liquid CO2 becomes saturated and the system reach equilibrium which causes the increment of the oil yield (Kumoro & Hasan, 2007). The fresh CO2 flowed to the sample matrix which could improve the Table 2 Regression coefficient, standard error, t and p value of CCD for P. macrocarpa seed oil extraction. Term
Coefficients
Standard error
t
pa
β0 X1 X2 X3 X21 X22 X23 X1*X2 X1*X3 X2*X3 R2 Adj. R2 Pred. R2
−117.636 0.280 6.750 5.546 −0.001 −0.076 −0.289 −0.065 −0.024 −0.097 0.9997 0.9994 0.9982
6.7695 0.2045 0.1202 1.3129 0.0014 0.0014 0.1426 0.0008 0.0083 0.0083
−17.377 1.369 56.162 4.224 −0.450 −53.479 −2.029 −2.407 −2.346 −1.377
0.000 0.204 0.000 0.002 0.663 0.000 0.073 0.039 0.044 0.202
a
p b 0.01 highly significant; 0.01 b p b 0.05 significant; p N 0.05 not significant.
50 Yield
40 4 42
30
36 30
60
70
80
Temperature
Pressure
24
(b)
48 Yield
47 46
5
45
4 60
70
80
Temperature
Flow rate
3
(c)
50 Yield
40 5
30 4 24
30 Pressure
36
42
Flow rate
3
Fig. 2. Response surface for the oil yield as a function of (a) temperature and pressure at a constant CO2 flow rate of 4 ml/min, (b) temperature and CO2 flow rate at a constant pressure of 35 MPa and (c) pressure and CO2 flow rate at a constant temperature of 70 °C.
interaction of sample with solvent as well as the extraction rate. By introducing more scCO2 with sample, the high CO2 flow rate can also accelerate the extraction. Fig. 2(c) describes the effect of pressure and CO2 flow rate on the extracted oil yield at the constant temperature of 70 °C. Increment of pressure and CO2 flow rate caused the increment of yield with the help of temperature, as shown in Fig. 2a and b. At low pressure, when the density of scCO2 was low, the changes of CO2 flow rate did not result any changes of the oil yield. But at high pressure, changes on the yield were observed with CO2 flow rate. In contrast, pressure caused the improvement of the oil yield at both low and high CO2 flow rates but improvement was greater at high flow rate than at low CO2 flow rate. The optimum extraction process parameters for the extraction of oil from P. macrocarpa within the given ranges were 72 °C, 42 MPa and CO2 flow rate of 4.5 ml/min. The oil yield in this optimized condition was obtained 52.9 g per 100 g of sample on dry weight basis which was 95.6%
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
of the oil extracted by conventional solvent extraction method. The oil yield obtained from the solvent extraction at optimized condition (72.2 °C, 8.4 h and solvent to feed ratio of 10.9 ml/g) was 55.32 g per 100 g of sample (Azmir et al., 2014). The amount of oil was extracted using scCO2 method in a shorter period of time (3 h) was similar to the amount extracted by solvent extraction process that required longer extraction time (8.4 h). Moreover, scCO2 does not use organic solvent thus makes the product green, environmentally friendly and healthy. Therefore, scCO2 can be used as effective alternative of conventional solvent extraction method for the extraction of oil from P. macrocarpa seed in an aspect of green and friendly method.
5
Predicted
R 2 = 0.999962
3.2. Analysis of ANN model
3.3. Analysis of fatty acid composition The fatty acid in terms of triglyceride constituents in P. macrocarpa seed oil extracted using scCO2 and conventional method are shown in Table 3. A total of five main fatty acids were identified from the scCO2 and solvent extracted oil by GC–MS analysis whereas three additional fatty acids (tetradecanoic acid 0.08%, trans-13-octadecanoic acid 0.22% and nonadecanoic acid 0.19%) and few other non-polar compounds (in negligible amount) were detected in the solvent extracted oil only (Azmir et al., 2014). The conventional solvent extraction is not a selective method which may be the possible reason for extracting some nonpolar compounds other than fatty acids but in negligible amount. The amount of palmitic acid was higher in scCO2 extract (20.52 ± 0.09%) than the conventional extract (15 ± 0.12%). Oleic acid was found to be the highest amount of 43.84 ± 0.52% among all the fatty acids in both extracts (43.56 ± 0.08% with conventional method). The linoleic acid was also in high amount in the extracted oil (29.18 ± 0.65%)
Observed
(a)
Observed
(b)
Predicted
R 2=0.999947
R 2=0.999485
Predicted
In this study, the feed-forward neural network was used because of the non-linear relationship between input parameters and the output and the complex properties of the selected process. The number of neurons, 6, in the hidden layer was selected by trial-and-error method where several networks were constructed and the best was selected based on the accuracy of the predictions in the testing phase. The mean square error (MSE) and R2 of the predicted and actual values were 0.0176 and 0.99 for training data set, 0.0232 and 0.99 for validation data set, and 0.0185 and 0.99 for all data set, respectively. These results prove the high predicting quality of constructed ANN model. The experimental yield and predicted yield by ANN constructed model were presented in Table 1. Fig. 3(a, b and c) shows the R2 values of the entire data set, the training and validation data set, respectively. For comparison between RSM and ANN, the same data set used in RSM was applied for the prediction of the yield in ANN. The R2 of the predicted data obtained from both RSM and ANN were the same (0.99). The R2 alone cannot measure the model accuracy for which the use of AAD analysis is necessary. AAD is a direct method for describing the deviations. A better accuracy can be evaluated using both R2 and AAD. The R2 found to be closer to 1.0 for both RSM and ANN. The AAD value for the ANN model was 0.25% that was lower compared to that of AAD (0.31%) for the RSM. The result describes that both models used in this study were good enough for prediction where the predictive quality of ANN was found to be better than that of RSM for the nonlinear data of extracted oil yield from P. macrocarpa seed. This result is comparable with the studies conducted by other researchers. The earlier study conducted by Baş and Boyaci (2007) showed ANN as a better technique than RSM for the estimation capabilities of biochemical reaction. The superiority of ANN in predicting the non-linear behavior over RSM was reported by Desai, Survase, Saudagar, Lele, and Singhal (2008) for fermentation system. Cheok, Chin, Yusof, Talib, and Law (2012) also found ANN as better predictive model for total phenolic content extraction compared to RSM with R2 of 0.94 and 0.89, and AAD of 4.01% and 5.37%, respectively.
Observed
(c)
Fig. 3. The scatter plots of predicted versus actual values of ANN model for (a) entire data set, (b) training data set and (c) validation data set.
and the stearic acid was in moderate amount (5.86 ± 0.09%) that was slightly higher than the stearic acid (4.11 ± 0.10%), extracted using n-hexane. According to Li et al. (2014), different extraction methods not only lead to different oil yields but also affect the fatty acid profile to a large extent. The solubility of solutes into scCO2 depends Table 3 Fatty acid constituents (% as methyl ester) of P. macrocarpa seed oil determined by GC–MS. Fatty acids
scCO2 extraction (%) (mean ± sd)
Conventional extractiona (%) (mean ± sd)
Palmitic acid (16:0) Linoleic acid (18:2) Oleic acid (18:1) Stearic acid (18:0) Gondoic acid (20:1) Others
20.52 ± 0.09 29.18 ± 0.65 43.84 ± 0.52 5.86 ± 0.09 0.70 ± 0.24 NA
15.00 36.25 43.56 04.11 00.32 00.76
a
± ± ± ± ± ±
0.12 0.23 0.08 0.10 0.07 0.16
Azmir et al., 2014.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
6
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
on the chemical structure such as molecular weight, chain length, and double bonds of the solute (Brunner & Machadob, 2012). The solubility of fatty acids in scCO2 decreases with an increase of carbon chain length as well as with an increase in the number of double bonds (Brunner & Machadob, 2012; Maheshwari, Nikolov, White, & Hartel, 1992). In scCO2 extraction, low molecular weight compound was being extracted rapidly in the initial part of extraction compared to high molecular weight compounds. With increasing temperature, the solubility of palmitic acid increases the most while oleic and linoleic acids increase the least (Maheshwari et al., 1992). The observed difference between the amount of scCO2 and solvent extracted palmitic acid might be due to the complete extraction of palmitic acid as it is a short chain fatty acid (low molecular weight) than other identified fatty acids and for temperature treatment. The GC–MS analysis of scCO2 extraction was done on FAMEs obtained from optimized condition. The optimized condition was based on the total oil yield rather than a particular fatty acid. The obtained linoleic acid in scCO2 was lower than solvent extraction possibly for the lower solubility of linoleic acid into scCO2 because of longer chain and double bond. Linoleic acid has been extracted in a later part of the extraction process and might have incomplete extraction in 3 h extraction period which ultimately lowered the percentage of linoleic acid in the total oil. The variation of the amount of co-extracted non-triglycerides (e.g. fat soluble vitamins, phospholipids) in the oil could also be responsible for the differences of fatty acid compositions of different extraction methods. The difference of fatty acids content for different extraction methods (scCO2 and solvent extraction) has been observed in several reports (Carvalho, Galvão, Barros, Conceição, & Sousa, 2012; Majdi, Barzegar, Jabbari, & AghaAlikhani, 2012; Wang, Sun, Chen, Qian, & Xu, 2011). A long chain fatty acid, gondoic acid (11-eicosanoic acid) was found in small quantities in the extracted oil (0.70 ± 0.24%). The total amount of monounsaturated fatty acids (73.62 ± 1.41%) in the oil was higher than that of the total saturated fatty acids (26.38 ± 0.18%). Thus the P. macrocarpa seed contains high amount of unsaturated fatty acids which could be a valuable oil after further analysis. 4. Conclusion The SFE was employed to extract oil from P. macrocarpa seed and the extraction pressure and flow rate of CO2 were found to have significant effect on oil yield. In the optimum conditions of the parameters obtained from CCD of RSM, the oil yield was 52.9 g per 100 g of sample on dry weight basis. The RSM and ANN were successfully applied to predict the oil yield as their R2 values were closer to 1 (0.99) and AAD of both models was also acceptable (0.31% for RSM and 0.25% for ANN). ANN showed better predictive efficiency compared with RSM. More than 50% oil was found from P. macrocarpa seed where the total of 73.72% was unsaturated fatty acids. In the unsaturated fatty acids, total of 44.54% monounsaturated and 29.18% polyunsaturated fatty acids were found. P. macrocarpa seed which is still considered as wild plant seed can be a promising source of valuable oil after proper processing. Acknowledgment This research was supported by the Endowment type-B grand from International Islamic University Malaysia (IIUM), Malaysia (Grant no: EDW B 13-063-0948). References Abbasi, H., Rezaei, K., & Rashidi, L. (2008). Extraction of essential oils from the seeds of pomegranate using organic solvents and supercritical CO2. Journal of the American Oil Chemists' Society, 85(1), 83–89. Ahnert, S. E., Travençolo, B.A., & da Fontoura Costa, L. (2009). Connectivity and dynamics of neuronal networks as defined by the shape of individual neurons. New Journal of Physics, 11(10), 103053.
Ali, R. B., Atangwho, I. J., Kaur, N., Abraika, O. S., Ahmad, M., Mahmud, R., et al. (2012). Bioassay-guided antidiabetic study of Phaleria macrocarpa fruit extract. Molecules, 17(5), 4986–5002. Azmir, J., Zaidul, I. S. M., Rahman, M. M., Sharif, K. M., Mohamed, A., Sahena, F., et al. (2013). Techniques for extraction of bioactive compounds from plant materials: A review. Journal of Food Engineering, 117(4), 426–436. Azmir, J., Zaidul, I. S. M., Rahman, M. M., Sharif, K. M., Sahena, F., Jahurul, M. H. A., et al. (2014). Optimization of oil yield of Phaleria macrocarpa seed using response surface methodology and its fatty acids constituents. Industrial Crops and Products, 52, 405–412. Baş, D., & Boyaci, İ. H. (2007). Modeling and optimization II: Comparison of estimation capabilities of response surface methodology with artificial neural networks in a biochemical reaction. Journal of Food Engineering, 78(3), 846–854. Brunner, G., & Machadob, N. T. (2012). Process design methodology for fractionation of fatty acids from palm fatty acid distillates in countercurrent packed columns with supercritical CO2. The Journal of Supercritical Fluids, 66(2012), 96–110. Carvalho, R. H. R., Galvão, E. L., Barros, J. A.C., Conceição, M. M., & Sousa, E. M. B.D. (2012). Extraction, fatty acid profile and antioxidant activity of sesame extract (Sesamum indicum L.). Brazilian Journal of Chemical Engineering, 29(2), 409–420. Cheok, C. Y., Chin, N. L., Yusof, Y. A., Talib, R. A., & Law, C. L. (2012). Optimization of total phenolic content extracted from Garcinia mangostana Linn. hull using response surface methodology versus artificial neural network. Industrial Crops and Products, 40, 247–253. Desai, K. M., Survase, S. A., Saudagar, P.S., Lele, S. S., & Singhal, R. S. (2008). Comparison of artificial neural network (ANN) and response surface methodology (RSM) in fermentation media optimization: Case study of fermentative production of scleroglucan. Biochemical Engineering Journal, 41(3), 266–273. Elkanzi, E. M., & Singh, H. (2001). Extraction of used lubricating oils with supercritical carbon dioxide. Arabian Journal for Science and Engineering, 26(2), 11–23. Erzin, Y., Rao, B. H., & Singh, D. N. (2008). Artificial neural network models for predicting soil thermal resistivity. International Journal of Thermal Sciences, 47(10), 1347–1358. Hendra, R., Ahmad, S., Oskoueian, E., Sukari, A., & Shukor, M. Y. (2011). Antioxidant, antiinflammatory and cytotoxicity of Phaleria macrocarpa (Boerl.) Scheff fruit. BMC Complementary and Alternative Medicine, 11(1), 110. Hussain, M., Bedi, J. S., & Singh, H. (1992, September). Determining number of neurons in hidden layers for binary error correcting codes. In Steven K. Rogers (Ed.), Aerospace Sensing (pp. 1015–1022). International Society for Optics and Photonics. Ibañez, E., Herrero, M., Mendiola, J. A., & Castro-Puyana, M. (2012). Extraction and characterization of bioactive compounds with health benefits from marine resources: Macro and micro algae, cyanobacteria, and invertebrates. In M. Hayes (Ed.), Marine bioactive compounds: Sources, characterization and applications (pp. 55–98). Springer. Ibrić, S., Djuriš, J., Parojčić, J., & Djurić, Z. (2012). Artificial neural networks in evaluation and optimization of modified release solid dosage forms. Pharmaceutics, 4(4), 531–550. Ichihara, K. I., & Fukubayashi, Y. (2010). Preparation of fatty acid methyl esters for gas– liquid chromatography. Journal of Lipid Research, 51(3), 635–640. Jorjani, E., Chehreh Chelgani, S., & Mesroghli, S. H. (2008). Application of artificial neural networks to predict chemical desulfurization of Tabas coal. Fuel, 87(12), 2727–2734. Kandhro, A., Sherazi, S. T. H., Mahesar, S. A., Bhanger, M. I., Talpur, M. Y., & Latif, Y. (2010). Variation in fatty acids composition including trans fat in different brands of potato chips by GC–MS. Pakistan Journal of Analytical and Environmental Chemistry, 11(1), 36–41. Kumoro, A.C., & Hasan, M. (2007). Supercritical carbon dioxide extraction of andrographolide from Andrographis paniculata: Effect of the solvent flow rate, pressure, and temperature. Chinese Journal of Chemical Engineering, 15(6), 877–883. Lang, Q., & Wai, C. M. (2001). Supercritical fluid extraction in herbal and natural product studies—A practical review. Talanta, 53(4), 771–782. Li, Y., Naghdi, F. G., Garg, S., Adarme-Vega, T. C., Thurecht, K. J., Ghafor, W. A., et al. (2014). A comparative study: The impact of different lipid extraction methods on current microalgal lipid research. Microbial Cell Factories, 13(1), 14. Maheshwari, P., Nikolov, Z. L., White, T. M., & Hartel, R. (1992). Solubility of fatty acids in supercritical carbon dioxide. Journal of the American Oil Chemists' Society, 69(11), 1069–1076. Majdi, S., Barzegar, M., Jabbari, A., & AghaAlikhani, M. (2012). Supercritical fluid extraction of tobacco seed oil and its comparison with solvent extraction methods. Journal of Agricultural Science and Technology, 14(5), 1053–1065. Mandal, S., Sivaprasad, P. V., Venugopal, S., & Murthy, K. P. N. (2009). Artificial neural network modeling to evaluate and predict the deformation behavior of stainless steel type AISI 304 L during hot torsion. Applied Soft Computing, 9(1), 237–244. Raymond, H. M., & Montgomery, D. C. (2002). Response surface methodology: Process and product optimization using designed experiment (2nd ed.). USA: A Wiley- Interscience Publication. Scalia, S., Giuffreda, L., & Pallado, P. (1999). Analytical and preparative supercritical fluid extraction of chamomile flowers and its comparison with conventional methods. Journal of Pharmaceutical and Biomedical Analysis, 21(3), 549–558. Sharif, K. M., Rahman, M. M., Azmir, J., Mohamed, A., Jahurul, M. H. A., Sahena, F., et al. (2014). Experimental design of supercritical fluid extraction–A review. Journal of Food Engineering, 124, 105–116. Susilawati, S., Matsjeh, S., Pranowo, H. D., & Anwar, C. (2011). Antioxidant activity of 2,6,4′-trihydroxy-4-methoxy benzophenone from ethyl acetate extract of leaves of Mahkota dewa (Phaleria macrocarpa (Scheff.) Boerl.). Indonesian Journal of Chemistry, 11(2), 180–185.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx Temelli, F., & Güçlü-Üstündağ, Ö. (2005). Supercritical technologies for further processing of edible oils. Bailey's industrial oil and fat products. John Wiley & Sons, Inc, http://dx. doi.org/10.1002/047167849X. Wang, H., Liu, Y., Wei, S., & Yan, Z. (2012). Application of response surface methodology to optimise supercritical carbon dioxide extraction of essential oil from Cyperus rotundus Linn. Food Chemistry, 132(1), 582–587.
7
Wang, Y., Sun, D., Chen, H., Qian, L., & Xu, P. (2011). Fatty acid composition and antioxidant activity of tea (Camellia sinensis L.) seed oil extracted by optimized supercritical carbon dioxide. International Journal of Molecular Sciences, 12(11), 7708–7719. Zhang, Y. B., Xu, X. J., & Liu, H. M. (2006). Chemical constituents from Mahkota dewa. Journal of Asian Natural Products Research, 8(1–2), 119–123.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
Contents lists available at ScienceDirect
Food Research International journal homepage: www.elsevier.com/locate/foodres
Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed J. Azmir a, I.S.M. Zaidul a,⁎, K.M. Sharif a, M.S. Uddin a, M.H.A. Jahurul b, S. Jinap c, Parvaneh Hajeb c, A. Mohamed d a
Faculty of Pharmacy, International Islamic University Malaysia, Kuantan Campus, 25200 Kuantan, Pahang, Malaysia Department of Food Science and Nutrition, Faculty of Applied Sciences, UCSI University, 56000 Kuala Lumpur, Malaysia Food Safety Research Centre (FOSREC), Faculty of Food Science and Technology, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia d Faculty of Pharmacy, Cyberjaya University College of Medical Sciences, 63000 Cyberjaya, Malaysia b c
a r t i c l e
i n f o
Article history: Received 24 April 2014 Received in revised form 22 June 2014 Accepted 28 June 2014 Available online xxxx Keywords: Phaleria macrocarpa seed oil Fatty acids Response surface methodology Artificial neural network
a b s t r a c t Good quality oil with high unsaturated fatty acids was found in the seed of a medicinal plant Phaleria macrocarpa (Mahkota dewa). Different parts especially fruit flesh of this plant are being traditionally used as important folk medicine whereas seed of this plant is usually neglected. In this study, the oil was extracted from P. macrocarpa seed using supercritical carbon dioxide. The extraction parameters were optimized by central composite design (CCD) of response surface methodology (RSM). Due to the non-linearity of the extraction process, artificial neural network (ANN) was also applied for predicting the oil yield. The optimum conditions obtained from RSM were 72 °C, 42 MPa and 4.5 ml/min CO2 flow rate where the oil yield was 52.9 g per 100 g of dry sample and coefficient of determination (R2) was 0.99. The ANN and RSM prediction showed similar R2 of 0.99 and ANN has lower average absolute deviation (AAD) of 0.25% compared to RSM (AAD of 0.31%). Five fatty acids were identified by gas chromatography–mass spectroscopy (GC–MS) analysis of the oil. The amount of oleic acid (18:1) was found to be highest (43.56%) among all the fatty acids. The total unsaturated fatty acid was 73.62% and saturated fatty acid was 26.38% in the P. macrocarpa seed oil. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Phaleria macrocarpa (local name: Mahkota dewa), member of Thymelaeaceae family has been traditionally treated as an important medicinal plant for centuries in Malaysia and Indonesia. Mahkota dewa means “God's Crown” meaning it descends from heaven to help mankind. Leaves, stem, fruit and seed of this plant have been examined extensively for their biological and pharmacological potencies (Ali et al., 2012; Hendra, Ahmad, Oskoueian, Sukari, & Shukor, 2011; Susilawati, Matsjeh, Pranowo, & Anwar, 2011; Zhang, Xu, & Liu, 2006). Apart from bioactive compounds responsible for pharmacological activities, the seed of this plant contains high amount of valuable oil reported by Azmir et al. (2014). Solvent extraction has been commonly used for the extraction of oil from plant matrix. It has, however, some inherent drawbacks; high heat is essential for the distillation of the oil, the oil contains residual solvents, and rancidity occurs during the separation process because the oil is oxidatively unstable (Scalia, Giuffreda, & Pallado, 1999). For the extraction of oil, supercritical carbon dioxide (scCO2) attracts considerable attention as a promising alternative to conventional organic ⁎ Corresponding author. Tel.: +60 9 570 4841; fax: +60 9 571 6775. E-mail address: [email protected] (I.S.M. Zaidul).
solvent extraction. Supercritical carbon dioxide has high extraction efficiency of oil and it is non-toxic, non-flammable, environmentally friendly, non-explosive, cost effective, less laborious, available and easy to remove from the extract thus eliminates post processing steps (Azmir et al., 2013; Lang & Wai, 2001). The extraction efficiency of scCO2 can be influenced by many factors such as extraction temperature and pressure, particle size and moisture content of feed material, time of extraction and flow rate of CO2 (Ibañez, Herrero, Mendiola, & Castro-Puyana, 2012; Temelli & Güçlü-Üstündağ, 2005). Response surface methodology (RSM) is a very effective technique to optimize significant operating parameters as well as to determine the effect of individual input variable and the interaction of variables for maximizing the benefits (Raymond & Montgomery, 2002; Wang, Liu, Wei, & Yan, 2012). Optimum and valid output can be produced with minimum effort, less time and resources by RSM as it requires minimum experimental runs for optimizing multiple independent variables (Sharif et al., 2014). A polynomial equation is used to describe and predict the response variable in RSM. On the other hand, artificial neural network (ANN) is very helpful to predict the response for non-linear system like extraction. ANN is a simplified model of the structure of a biological network (Mandal, Sivaprasad, Venugopal, & Murthy, 2009) and an artificial neuron is its main processing element. The neuron receives input data from various sources, combines the
http://dx.doi.org/10.1016/j.foodres.2014.06.049 0963-9969/© 2014 Elsevier Ltd. All rights reserved.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
Pressure Controller
Valve
Cooler
Temperature Controller
Back Pressure Regulator
CO2 Pump
Extraction Vessel
data, generally performs a nonlinear operation on the result and then gives output as the final result (Baş & Boyaci, 2007; Ibrić, Djuriš, Parojčić, & Djurić, 2012). The main advantage of ANN is that it does not need any mathematical modeling (Mandal et al., 2009) like linear and polynomial regressions based modeling. The aim of this study was to extract the P. macrocarpa seed oil using scCO2. RSM and ANN were employed to model the extraction process and to compare the prediction quality. RSM with central composite design (CCD) was employed to optimize the process parameters including extraction temperature, pressure and CO2 flow rate. The analysis of fatty acids in terms of triglycerides composition of the extracted P. macrocarpa seed oil was also aimed in this study.
Valve
2
2. Method and materials Collector
2.1. Materials
(a)
CO2 Cylinder
The fresh fruits of P. macrocarpa were collected from a local supplier in Kuantan, Malaysia. The sample specimen was deposited in herbarium of International Islamic University Malaysia (IIUM), Kuantan, Malaysia. All reagents (analytical and HPLC grade) were purchased from Merck Ltd. (Darmstadt, Germany). The CO2 gas was purchased from Fuelink Marketing Sdn. Bhd. (Selangor, Malaysia).
Hidden layer
2.2. Extraction of oil by conventional method
Temperature Oil from P. macrocarpa seed was extracted by solvent extraction technique using n-hexane at optimum conditions of 72 °C for 8.4 h and solvent-to-feed ratio of 10.9 ml/g as reported in our previous study (Azmir et al., 2014). The total oil yields were determined in percentage of 100 g of ground seed sample on dry weight basis.
CO2 flow rate
2.3. Supercritical carbon dioxide (scCO2) extraction The schematic diagram of scCO2 apparatus used in this study is shown in Fig. 1a. The apparatus has a CO2 pump (model PU-1580 and model PU-2080, Jasco Corporation, Tokyo, Japan) and the pump was fitted with a cooling jacket to cool the CO2 supplied from the tank. Ethylene glycol–deionized water mixture (50:50, v/v) was circulated through the cooling jacket using a low-temperature bath circulator (model 631D, Tech-Lab manufacturing Sdn. Bhd., Selangor, Malaysia) to chill coolant down to −6.5 °C. For every experiments, 10 g of sample was loaded into a 50 ml extraction vessel (model EV-3, Jasco Corporation, Tokyo), and the vessel was maintained at the desired temperature. A back-pressure regulator (BPR) (model BP-1580-81, Jasco Corporation, Tokyo, Japan) was used to control the extraction pressure. The seed oil was separated from the supercritical fluid by pressure reduction through an expansion valve. A 20 ml blue cap schott bottle was used for collecting the oil. The extraction process was continued for 3 h. The total yields were determined as percentage on dry weight basis of sample as described in the Eq. (1). Yield ð%Þ ¼
Massextracted oil 100: Massseed powder
Yield
Pressure
ð1Þ
2.4. Response surface methodology (RSM) RSM with CCD was employed to determine the best combinations of variables for obtaining high extraction yield in scCO2 process. Three parameters, temperature (X1), pressure (X2), and flow rate of CO2 (X3) were considered as independent variables. The selection of variables and their level (temperature, 60 to 80 °C; pressure, 25 to 45 MPa and CO2 flow rate, 3 to 5 ml/min) was based on the literature review, and trial-error (data not shown). The formulated experimental design was required 20 experimental runs for this experiment. For estimation of
(b) Fig. 1. (a) Schematic diagram of the SFE apparatus and (b) Schematic structure of studied ANN model.
“pure error” six replicates were performed at the central point (0,0,0) of the design. The run order and the combination of variables are shown in Table 1. A second-order polynomial regression equation was used for predicting the response variable (Y) as Eq. (2). Y ¼ β0 þ
X
βi X i þ
X
2
βii X i þ
XX
βij X i X j
ð2Þ
Where, Y is the response variable, β0 is a constant, and βi, βii and βij represent the linear, quadratic and interactive coefficients, respectively. Xi and Xj are the independent variables. The Minitab software (version 16) was used for multiple regression analysis, analysis of variance (ANOVA), and determination of coefficient R2 which measures the goodness of fit of regression model. The t-value was also included for the estimated coefficients and associated probabilities. The total error criteria with a confidence level of 95.0% were the basis for test of statistical significance. 2.5. Artificial neural network (ANN) The 20 experimental data used for RSM were also used for ANN to predict the response output of oil yield. The total experimental data was randomly split into two data sets: 17 for training and 3 for validation. The training data sets were applied for computing the network
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
3
Table 1 The experimental and predicted oil yield obtained from RSM and ANN. Temperature (°C)
Pressure (MPa)
CO2 flow rate (ml/min)
Experimental yield (g) (Mean ± sd)
Predicted yield (g) (RSM)
Predicted yield (g) (ANN)
80 70 80 70 60 80 70 60 60 60 70 80 70 70 70 60 70 70 70 80
25 35 25 35 45 45 35 45 25 25 35 45 35 35 45 35 25 35 35 35
3 4 5 4 3 5 4 5 5 3 4 3 5 4 4 4 4 4 3 4
26.35 47.08 28.91 47.56 50.00 53.05 47.49 52.88 28.14 24.61 47.23 50.75 48.86 47.18 51.87 46.90 27.67 47.33 45.35 47.76
26.28 47.32 29.08 47.32 49.84 53.03 47.32 52.96 28.21 24.64 47.32 50.69 48.54 47.35 52.02 46.86 27.44 47.35 45.59 47.72
26.32 47.34 28.91 47.34 50.07 53.14 47.34 52.90 28.21 24.70 47.34 50.77 48.84 47.34 51.66 46.73 27.53 47.34 45.59 47.66
parameters and the validation data sets were used for ensuring the robustness testing of the network parameters. A feed forward ANN which uses back-propagation algorithm, based on a multi-layer perceptron (MLP) was applied using Peltarian Synapse (Version 1.6.0, the Peltarion Corporation, Stockholm, Sweden) for the modeling. MLP with three layers, the most popular ANN (Hussain, Bedi, & Singh, 1992; Jorjani, Chehreh Chelgani, & Mesroghli, 2008) which was consisted of an input layer, hidden layers and an output layer was used for this experiment. Extraction temperature, pressure and CO2 flow rate were the input factors and yield of oil was the output. The structure of ANN is shown in Fig. 1b. The transfer function (weight layer) that expresses an internal activation level is the constituent of neuron (Ahnert, Travençolo, & da Fontoura Costa, 2009). Sigmoidal function was used as transfer function to determine the output from a neuron by transforming its input. The input and output data were normalized to the range between 0 and 1 to avoid large magnitude of the data using Eq. (3) (Erzin, Rao, & Singh, 2008). Xnorm ¼
X−Xmin Xmax −Xmin
ð3Þ
Where, Xnorm is the normalized value, X is the actual value, Xmax is the maximum value and Xmin is the minimum value. The mean squared error (MSE) and R2 were used to test the performance of the developed network by Eqs. (4) and (5), respectively. MSE ¼
1 Xn 2 ðYi −Ydi Þ i¼1 n Xn
ð4Þ
2
ðYi −Ydi Þ R ¼ 1−Xni¼1 2 ðYdi −Ym Þ i¼1 2
ð5Þ
Where, n is the number of points, Yi is the predicted value obtained from the neural network model, Ydi is the actual value, and Ym is the average of the actual values. The prediction capabilities of the two techniques, RSM and ANN were compared by the R2 value and absolute average deviation (AAD). The AAD was calculated using Eq. (6) (Baş & Boyaci, 2007): AAD ¼
nhX
p i¼1
i o P 100: Yi; exp −Yi;cal Yi; exp
=
=
ð6Þ
± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±
0.04 0.08 0.03 0.20 0.23 0.12 0.15 0.08 0.06 0.06 0.03 0.07 0.06 0.08 0.02 0.20 0.07 0.15 0.10 0.16
Where, Yi,exp and Yi,cal are the experimental and calculated responses, respectively. P is the number of the experimental run. 2.6. Fatty acid composition analysis by GC–MS Fatty acids were derivatized properly before the GC–MS analysis and the derivatization was done according to the method of Ichihara and Fukubayashi (2010) with some modifications. The oil sample was taken for experiment after being melted at 60–70 °C and homogenized thoroughly. Commercial concentrated HCl (35% w/w) was diluted with 41.5 ml methanol to make 50 ml 8.0% (w/v) HCl reagent. The HCl reagent was stored in a refrigerator until use. A black screw cap glass test tube was used to take exactly 100 μl of oil sample and the sample was dissolved in 0.2 ml toluene. After that, 1.5 ml methanol and 0.3 ml of HCl reagent were added to the oil solution. The tube was vortexed and then heated at 95 °C for 90 min for rapid methylation. After cooling the oil solution at room temperature, 1 ml hexane and 1 ml water were added for the extraction of fatty acid methyl esters (FAMEs). The tube was vortexed again and the hexane layer with FAMEs was separated. For the composition analysis of FAMEs, an Agilent 6890 N gas chromatography coupled with Agilent MS-5973 mass selective detector (Agilent Technologies; USA, serial no. US14113031) was used. The prepared FAMEs were injected onto the HP-5MS column (30 m dimension, 0.25 mm i.d, 0.25 μm film thickness) using an Agilent autosampler 7683 series injector and the split ratio was 50:1. The initial oven temperature was maintained at 150 ºC for 2 min and increased at a rate of 4 °C/min up to 230 °C, then kept at 230 °C for 5 min. The injector and detector temperatures were held at 240 °C and 260 °C, respectively. Helium was the carrier gas used at a flow rate of 0.8 ml/min. A 70 eV electron impact (EI) in a 50–550 m/z scan range was applied for the operation of mass spectrometer (Kandhro et al., 2008). For the analysis, ChemStation integrated software (Agilent Technologies) was used. The fatty acids were tentatively identified by matching the mass spectra with NIST mass spectral database (NIST08). 3. Result and discussion 3.1. Analysis of RSM model The experimental oil yield for different combinations of input variables obtained from CCD is presented in Table 1. The input and response variables were analyzed to obtain the predicted response for
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
4
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
oil yield by the second-order polynomial equation as presented in Eq. (7).
(a)
2
Y ¼ 0:280 X 1 þ 6:750 X 2 þ 5:546 X 3 −0:001 X 1 −0:076 2 2 ð7Þ X 2 −0:289 X 3 −0:002 X 1 X 2 −0:019 X 1 X 3 −0:011 X 2 X 3 −117:636 The multiple regression coefficients of the model calculated by the least square technique are demonstrated in Table 2. The R2 of the model was 0.99 that indicates a good agreement between the experimental and predicted yield of oil. The lack of fit of this model was non-significant with the p value of 0.329 meaning that the second order polynomial equation was adequate for predicting the oil yield under all combinations of values of the factors. The linear and square effects of pressure were highly significant (p ≤ 0.01) where in both cases the effects of temperature were non-significant (p ≥ 0.05). The linear effect of CO2 flow rate was significant (p = 0.002) but the quadric effect was not significant (p = 0.073). The interaction between pressure and temperature and temperature and CO2 flow rate was significant (p ≤ 0.05) within the experimental ranges. The interactions of the studied factors on the oil yield are represented by three-dimensional response surface shown in Fig. 2. The interaction effect of temperature and pressure on the oil yield at a fixed CO2 flow rate of 4 ml/min is shown as surface plot in Fig. 2(a). At constant temperature, the yield increased significantly with pressure. Pressure showed positive linear effect at both low and high temperatures. On the other hand, temperature did not change the yield much neither at low pressure nor at high pressure. Density of scCO2 increases with pressure which improves the solubility and mass transfer of oil (Abbasi, Rezaei, & Rashidi, 2008). Fig. 2(b) represents the response surface and contour plot depicting the influence of temperature and CO2 flow rate on the oil yield at constant pressure (35 MPa). The flow rate of CO2 had a positive effect on the oil yield irrespective of the level of temperature whereas the temperature had positive effect only at low CO2 flow rate. At low flow rate, the increase of the temperature increased the yield of oil. When the flow rate reached to 5 ml/min, the effect of flow rate became more prominent on the oil yield that might suppressed the effect of the temperature. In contrast, the oil yield increased with CO2 flow rate in both, at high and at low temperatures. The possible explanation of this phenomenon is that the equilibrium state between solute and solvent could be favorable for the high mass transfer rate (Elkanzi & Singh, 2001; Kumoro & Hasan, 2007). At low CO2 flow rate, mass transfer resistance reduces the amount of oil transported to the CO2 where with the increase of CO2 flow rate the mass transfer resistance declines until liquid CO2 becomes saturated and the system reach equilibrium which causes the increment of the oil yield (Kumoro & Hasan, 2007). The fresh CO2 flowed to the sample matrix which could improve the Table 2 Regression coefficient, standard error, t and p value of CCD for P. macrocarpa seed oil extraction. Term
Coefficients
Standard error
t
pa
β0 X1 X2 X3 X21 X22 X23 X1*X2 X1*X3 X2*X3 R2 Adj. R2 Pred. R2
−117.636 0.280 6.750 5.546 −0.001 −0.076 −0.289 −0.065 −0.024 −0.097 0.9997 0.9994 0.9982
6.7695 0.2045 0.1202 1.3129 0.0014 0.0014 0.1426 0.0008 0.0083 0.0083
−17.377 1.369 56.162 4.224 −0.450 −53.479 −2.029 −2.407 −2.346 −1.377
0.000 0.204 0.000 0.002 0.663 0.000 0.073 0.039 0.044 0.202
a
p b 0.01 highly significant; 0.01 b p b 0.05 significant; p N 0.05 not significant.
50 Yield
40 4 42
30
36 30
60
70
80
Temperature
Pressure
24
(b)
48 Yield
47 46
5
45
4 60
70
80
Temperature
Flow rate
3
(c)
50 Yield
40 5
30 4 24
30 Pressure
36
42
Flow rate
3
Fig. 2. Response surface for the oil yield as a function of (a) temperature and pressure at a constant CO2 flow rate of 4 ml/min, (b) temperature and CO2 flow rate at a constant pressure of 35 MPa and (c) pressure and CO2 flow rate at a constant temperature of 70 °C.
interaction of sample with solvent as well as the extraction rate. By introducing more scCO2 with sample, the high CO2 flow rate can also accelerate the extraction. Fig. 2(c) describes the effect of pressure and CO2 flow rate on the extracted oil yield at the constant temperature of 70 °C. Increment of pressure and CO2 flow rate caused the increment of yield with the help of temperature, as shown in Fig. 2a and b. At low pressure, when the density of scCO2 was low, the changes of CO2 flow rate did not result any changes of the oil yield. But at high pressure, changes on the yield were observed with CO2 flow rate. In contrast, pressure caused the improvement of the oil yield at both low and high CO2 flow rates but improvement was greater at high flow rate than at low CO2 flow rate. The optimum extraction process parameters for the extraction of oil from P. macrocarpa within the given ranges were 72 °C, 42 MPa and CO2 flow rate of 4.5 ml/min. The oil yield in this optimized condition was obtained 52.9 g per 100 g of sample on dry weight basis which was 95.6%
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
of the oil extracted by conventional solvent extraction method. The oil yield obtained from the solvent extraction at optimized condition (72.2 °C, 8.4 h and solvent to feed ratio of 10.9 ml/g) was 55.32 g per 100 g of sample (Azmir et al., 2014). The amount of oil was extracted using scCO2 method in a shorter period of time (3 h) was similar to the amount extracted by solvent extraction process that required longer extraction time (8.4 h). Moreover, scCO2 does not use organic solvent thus makes the product green, environmentally friendly and healthy. Therefore, scCO2 can be used as effective alternative of conventional solvent extraction method for the extraction of oil from P. macrocarpa seed in an aspect of green and friendly method.
5
Predicted
R 2 = 0.999962
3.2. Analysis of ANN model
3.3. Analysis of fatty acid composition The fatty acid in terms of triglyceride constituents in P. macrocarpa seed oil extracted using scCO2 and conventional method are shown in Table 3. A total of five main fatty acids were identified from the scCO2 and solvent extracted oil by GC–MS analysis whereas three additional fatty acids (tetradecanoic acid 0.08%, trans-13-octadecanoic acid 0.22% and nonadecanoic acid 0.19%) and few other non-polar compounds (in negligible amount) were detected in the solvent extracted oil only (Azmir et al., 2014). The conventional solvent extraction is not a selective method which may be the possible reason for extracting some nonpolar compounds other than fatty acids but in negligible amount. The amount of palmitic acid was higher in scCO2 extract (20.52 ± 0.09%) than the conventional extract (15 ± 0.12%). Oleic acid was found to be the highest amount of 43.84 ± 0.52% among all the fatty acids in both extracts (43.56 ± 0.08% with conventional method). The linoleic acid was also in high amount in the extracted oil (29.18 ± 0.65%)
Observed
(a)
Observed
(b)
Predicted
R 2=0.999947
R 2=0.999485
Predicted
In this study, the feed-forward neural network was used because of the non-linear relationship between input parameters and the output and the complex properties of the selected process. The number of neurons, 6, in the hidden layer was selected by trial-and-error method where several networks were constructed and the best was selected based on the accuracy of the predictions in the testing phase. The mean square error (MSE) and R2 of the predicted and actual values were 0.0176 and 0.99 for training data set, 0.0232 and 0.99 for validation data set, and 0.0185 and 0.99 for all data set, respectively. These results prove the high predicting quality of constructed ANN model. The experimental yield and predicted yield by ANN constructed model were presented in Table 1. Fig. 3(a, b and c) shows the R2 values of the entire data set, the training and validation data set, respectively. For comparison between RSM and ANN, the same data set used in RSM was applied for the prediction of the yield in ANN. The R2 of the predicted data obtained from both RSM and ANN were the same (0.99). The R2 alone cannot measure the model accuracy for which the use of AAD analysis is necessary. AAD is a direct method for describing the deviations. A better accuracy can be evaluated using both R2 and AAD. The R2 found to be closer to 1.0 for both RSM and ANN. The AAD value for the ANN model was 0.25% that was lower compared to that of AAD (0.31%) for the RSM. The result describes that both models used in this study were good enough for prediction where the predictive quality of ANN was found to be better than that of RSM for the nonlinear data of extracted oil yield from P. macrocarpa seed. This result is comparable with the studies conducted by other researchers. The earlier study conducted by Baş and Boyaci (2007) showed ANN as a better technique than RSM for the estimation capabilities of biochemical reaction. The superiority of ANN in predicting the non-linear behavior over RSM was reported by Desai, Survase, Saudagar, Lele, and Singhal (2008) for fermentation system. Cheok, Chin, Yusof, Talib, and Law (2012) also found ANN as better predictive model for total phenolic content extraction compared to RSM with R2 of 0.94 and 0.89, and AAD of 4.01% and 5.37%, respectively.
Observed
(c)
Fig. 3. The scatter plots of predicted versus actual values of ANN model for (a) entire data set, (b) training data set and (c) validation data set.
and the stearic acid was in moderate amount (5.86 ± 0.09%) that was slightly higher than the stearic acid (4.11 ± 0.10%), extracted using n-hexane. According to Li et al. (2014), different extraction methods not only lead to different oil yields but also affect the fatty acid profile to a large extent. The solubility of solutes into scCO2 depends Table 3 Fatty acid constituents (% as methyl ester) of P. macrocarpa seed oil determined by GC–MS. Fatty acids
scCO2 extraction (%) (mean ± sd)
Conventional extractiona (%) (mean ± sd)
Palmitic acid (16:0) Linoleic acid (18:2) Oleic acid (18:1) Stearic acid (18:0) Gondoic acid (20:1) Others
20.52 ± 0.09 29.18 ± 0.65 43.84 ± 0.52 5.86 ± 0.09 0.70 ± 0.24 NA
15.00 36.25 43.56 04.11 00.32 00.76
a
± ± ± ± ± ±
0.12 0.23 0.08 0.10 0.07 0.16
Azmir et al., 2014.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
6
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx
on the chemical structure such as molecular weight, chain length, and double bonds of the solute (Brunner & Machadob, 2012). The solubility of fatty acids in scCO2 decreases with an increase of carbon chain length as well as with an increase in the number of double bonds (Brunner & Machadob, 2012; Maheshwari, Nikolov, White, & Hartel, 1992). In scCO2 extraction, low molecular weight compound was being extracted rapidly in the initial part of extraction compared to high molecular weight compounds. With increasing temperature, the solubility of palmitic acid increases the most while oleic and linoleic acids increase the least (Maheshwari et al., 1992). The observed difference between the amount of scCO2 and solvent extracted palmitic acid might be due to the complete extraction of palmitic acid as it is a short chain fatty acid (low molecular weight) than other identified fatty acids and for temperature treatment. The GC–MS analysis of scCO2 extraction was done on FAMEs obtained from optimized condition. The optimized condition was based on the total oil yield rather than a particular fatty acid. The obtained linoleic acid in scCO2 was lower than solvent extraction possibly for the lower solubility of linoleic acid into scCO2 because of longer chain and double bond. Linoleic acid has been extracted in a later part of the extraction process and might have incomplete extraction in 3 h extraction period which ultimately lowered the percentage of linoleic acid in the total oil. The variation of the amount of co-extracted non-triglycerides (e.g. fat soluble vitamins, phospholipids) in the oil could also be responsible for the differences of fatty acid compositions of different extraction methods. The difference of fatty acids content for different extraction methods (scCO2 and solvent extraction) has been observed in several reports (Carvalho, Galvão, Barros, Conceição, & Sousa, 2012; Majdi, Barzegar, Jabbari, & AghaAlikhani, 2012; Wang, Sun, Chen, Qian, & Xu, 2011). A long chain fatty acid, gondoic acid (11-eicosanoic acid) was found in small quantities in the extracted oil (0.70 ± 0.24%). The total amount of monounsaturated fatty acids (73.62 ± 1.41%) in the oil was higher than that of the total saturated fatty acids (26.38 ± 0.18%). Thus the P. macrocarpa seed contains high amount of unsaturated fatty acids which could be a valuable oil after further analysis. 4. Conclusion The SFE was employed to extract oil from P. macrocarpa seed and the extraction pressure and flow rate of CO2 were found to have significant effect on oil yield. In the optimum conditions of the parameters obtained from CCD of RSM, the oil yield was 52.9 g per 100 g of sample on dry weight basis. The RSM and ANN were successfully applied to predict the oil yield as their R2 values were closer to 1 (0.99) and AAD of both models was also acceptable (0.31% for RSM and 0.25% for ANN). ANN showed better predictive efficiency compared with RSM. More than 50% oil was found from P. macrocarpa seed where the total of 73.72% was unsaturated fatty acids. In the unsaturated fatty acids, total of 44.54% monounsaturated and 29.18% polyunsaturated fatty acids were found. P. macrocarpa seed which is still considered as wild plant seed can be a promising source of valuable oil after proper processing. Acknowledgment This research was supported by the Endowment type-B grand from International Islamic University Malaysia (IIUM), Malaysia (Grant no: EDW B 13-063-0948). References Abbasi, H., Rezaei, K., & Rashidi, L. (2008). Extraction of essential oils from the seeds of pomegranate using organic solvents and supercritical CO2. Journal of the American Oil Chemists' Society, 85(1), 83–89. Ahnert, S. E., Travençolo, B.A., & da Fontoura Costa, L. (2009). Connectivity and dynamics of neuronal networks as defined by the shape of individual neurons. New Journal of Physics, 11(10), 103053.
Ali, R. B., Atangwho, I. J., Kaur, N., Abraika, O. S., Ahmad, M., Mahmud, R., et al. (2012). Bioassay-guided antidiabetic study of Phaleria macrocarpa fruit extract. Molecules, 17(5), 4986–5002. Azmir, J., Zaidul, I. S. M., Rahman, M. M., Sharif, K. M., Mohamed, A., Sahena, F., et al. (2013). Techniques for extraction of bioactive compounds from plant materials: A review. Journal of Food Engineering, 117(4), 426–436. Azmir, J., Zaidul, I. S. M., Rahman, M. M., Sharif, K. M., Sahena, F., Jahurul, M. H. A., et al. (2014). Optimization of oil yield of Phaleria macrocarpa seed using response surface methodology and its fatty acids constituents. Industrial Crops and Products, 52, 405–412. Baş, D., & Boyaci, İ. H. (2007). Modeling and optimization II: Comparison of estimation capabilities of response surface methodology with artificial neural networks in a biochemical reaction. Journal of Food Engineering, 78(3), 846–854. Brunner, G., & Machadob, N. T. (2012). Process design methodology for fractionation of fatty acids from palm fatty acid distillates in countercurrent packed columns with supercritical CO2. The Journal of Supercritical Fluids, 66(2012), 96–110. Carvalho, R. H. R., Galvão, E. L., Barros, J. A.C., Conceição, M. M., & Sousa, E. M. B.D. (2012). Extraction, fatty acid profile and antioxidant activity of sesame extract (Sesamum indicum L.). Brazilian Journal of Chemical Engineering, 29(2), 409–420. Cheok, C. Y., Chin, N. L., Yusof, Y. A., Talib, R. A., & Law, C. L. (2012). Optimization of total phenolic content extracted from Garcinia mangostana Linn. hull using response surface methodology versus artificial neural network. Industrial Crops and Products, 40, 247–253. Desai, K. M., Survase, S. A., Saudagar, P.S., Lele, S. S., & Singhal, R. S. (2008). Comparison of artificial neural network (ANN) and response surface methodology (RSM) in fermentation media optimization: Case study of fermentative production of scleroglucan. Biochemical Engineering Journal, 41(3), 266–273. Elkanzi, E. M., & Singh, H. (2001). Extraction of used lubricating oils with supercritical carbon dioxide. Arabian Journal for Science and Engineering, 26(2), 11–23. Erzin, Y., Rao, B. H., & Singh, D. N. (2008). Artificial neural network models for predicting soil thermal resistivity. International Journal of Thermal Sciences, 47(10), 1347–1358. Hendra, R., Ahmad, S., Oskoueian, E., Sukari, A., & Shukor, M. Y. (2011). Antioxidant, antiinflammatory and cytotoxicity of Phaleria macrocarpa (Boerl.) Scheff fruit. BMC Complementary and Alternative Medicine, 11(1), 110. Hussain, M., Bedi, J. S., & Singh, H. (1992, September). Determining number of neurons in hidden layers for binary error correcting codes. In Steven K. Rogers (Ed.), Aerospace Sensing (pp. 1015–1022). International Society for Optics and Photonics. Ibañez, E., Herrero, M., Mendiola, J. A., & Castro-Puyana, M. (2012). Extraction and characterization of bioactive compounds with health benefits from marine resources: Macro and micro algae, cyanobacteria, and invertebrates. In M. Hayes (Ed.), Marine bioactive compounds: Sources, characterization and applications (pp. 55–98). Springer. Ibrić, S., Djuriš, J., Parojčić, J., & Djurić, Z. (2012). Artificial neural networks in evaluation and optimization of modified release solid dosage forms. Pharmaceutics, 4(4), 531–550. Ichihara, K. I., & Fukubayashi, Y. (2010). Preparation of fatty acid methyl esters for gas– liquid chromatography. Journal of Lipid Research, 51(3), 635–640. Jorjani, E., Chehreh Chelgani, S., & Mesroghli, S. H. (2008). Application of artificial neural networks to predict chemical desulfurization of Tabas coal. Fuel, 87(12), 2727–2734. Kandhro, A., Sherazi, S. T. H., Mahesar, S. A., Bhanger, M. I., Talpur, M. Y., & Latif, Y. (2010). Variation in fatty acids composition including trans fat in different brands of potato chips by GC–MS. Pakistan Journal of Analytical and Environmental Chemistry, 11(1), 36–41. Kumoro, A.C., & Hasan, M. (2007). Supercritical carbon dioxide extraction of andrographolide from Andrographis paniculata: Effect of the solvent flow rate, pressure, and temperature. Chinese Journal of Chemical Engineering, 15(6), 877–883. Lang, Q., & Wai, C. M. (2001). Supercritical fluid extraction in herbal and natural product studies—A practical review. Talanta, 53(4), 771–782. Li, Y., Naghdi, F. G., Garg, S., Adarme-Vega, T. C., Thurecht, K. J., Ghafor, W. A., et al. (2014). A comparative study: The impact of different lipid extraction methods on current microalgal lipid research. Microbial Cell Factories, 13(1), 14. Maheshwari, P., Nikolov, Z. L., White, T. M., & Hartel, R. (1992). Solubility of fatty acids in supercritical carbon dioxide. Journal of the American Oil Chemists' Society, 69(11), 1069–1076. Majdi, S., Barzegar, M., Jabbari, A., & AghaAlikhani, M. (2012). Supercritical fluid extraction of tobacco seed oil and its comparison with solvent extraction methods. Journal of Agricultural Science and Technology, 14(5), 1053–1065. Mandal, S., Sivaprasad, P. V., Venugopal, S., & Murthy, K. P. N. (2009). Artificial neural network modeling to evaluate and predict the deformation behavior of stainless steel type AISI 304 L during hot torsion. Applied Soft Computing, 9(1), 237–244. Raymond, H. M., & Montgomery, D. C. (2002). Response surface methodology: Process and product optimization using designed experiment (2nd ed.). USA: A Wiley- Interscience Publication. Scalia, S., Giuffreda, L., & Pallado, P. (1999). Analytical and preparative supercritical fluid extraction of chamomile flowers and its comparison with conventional methods. Journal of Pharmaceutical and Biomedical Analysis, 21(3), 549–558. Sharif, K. M., Rahman, M. M., Azmir, J., Mohamed, A., Jahurul, M. H. A., Sahena, F., et al. (2014). Experimental design of supercritical fluid extraction–A review. Journal of Food Engineering, 124, 105–116. Susilawati, S., Matsjeh, S., Pranowo, H. D., & Anwar, C. (2011). Antioxidant activity of 2,6,4′-trihydroxy-4-methoxy benzophenone from ethyl acetate extract of leaves of Mahkota dewa (Phaleria macrocarpa (Scheff.) Boerl.). Indonesian Journal of Chemistry, 11(2), 180–185.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049
J. Azmir et al. / Food Research International xxx (2014) xxx–xxx Temelli, F., & Güçlü-Üstündağ, Ö. (2005). Supercritical technologies for further processing of edible oils. Bailey's industrial oil and fat products. John Wiley & Sons, Inc, http://dx. doi.org/10.1002/047167849X. Wang, H., Liu, Y., Wei, S., & Yan, Z. (2012). Application of response surface methodology to optimise supercritical carbon dioxide extraction of essential oil from Cyperus rotundus Linn. Food Chemistry, 132(1), 582–587.
7
Wang, Y., Sun, D., Chen, H., Qian, L., & Xu, P. (2011). Fatty acid composition and antioxidant activity of tea (Camellia sinensis L.) seed oil extracted by optimized supercritical carbon dioxide. International Journal of Molecular Sciences, 12(11), 7708–7719. Zhang, Y. B., Xu, X. J., & Liu, H. M. (2006). Chemical constituents from Mahkota dewa. Journal of Asian Natural Products Research, 8(1–2), 119–123.
Please cite this article as: Azmir, J., et al., Supercritical carbon dioxide extraction of highly unsaturated oil from Phaleria macrocarpa seed, Food Research International (2014), http://dx.doi.org/10.1016/j.foodres.2014.06.049