Thermal inactivation of lipoxygenase in soya bean using superheated steam to produce low beany flavour soya milk

Abstract

Time and temperature parameters of superheated steam (SHS) treatment were optimised using response surface methodology (RSM) for specific lipoxygenase (LOX) activity in soya beans and crude protein content in soya milk. The optimal SHS treatment was obtained at 9.3 min and 119 °C. The predicted values of specific LOX activity and crude protein content by RSM were 0.0098 μmol/(min mg protein) and 3.2%, respectively. These values were experimentally verified to be 0.0081 ± 0.0002 μmol/(min mg protein) and 3.0 ± 0.1%, respectively. Sensory evaluation showed that the beany flavour of soya milk produced from SHS treated soya beans was significantly weaker (P < 0.05) than that of untreated soya beans. The results showed that the optimised SHS treatment could reduce the beany flavour in the soya milk significantly (P < 0.05) by reducing the specific LOX activity in the soybean, while ensuring the crude protein content in the soya milk complied with Malaysian Food Regulations 1985.

Introduction

Soya milk is nutritional and economical as the soya beans (Glycine max) are easily available locally in Asian countries. Soya food products, such as soya milk, tofu, and soya sauce are normally consumed daily by Asians (Wang et al. 1998). Soya milk has become an alternative to dairy milk for those who are lactose intolerance or allergic to milk proteins, as well as for those who avoid milk for other reasons, such as religion or diet (Imran 2007; Toro-Funes et al. 2015). In addition, soya milk is cholesterol-free, low in saturated fat, and high in phytochemicals, especially the isoflavone compounds (Wang et al. 2013). With the high isoflavone content, it can decrease low-density lipoprotein (LDL) and increase high-density lipoprotein (HDL) level. This helps to prevent hypercholesterolaemia and atherosclerosis (Asif and Acharya 2013). At the same time, consuming soya milk could help to reduce the risk of cardiovascular disease, osteoporosis, and to some extent cancers (Messina and Messina 2010; Messina 2016). According to Messina (2016), regular consumption of soya milk could enhance and protect consumers’ health.

The traditional orient processing of soya milk involves four main steps, which are (1) soaking the beans, (2) grinding in water, (3) filtering to remove the residue, and (4) cooking (Kwok and Niranjin 1995). The soya beans need to be soaked in water for overnight to soften the bean. The soaked soya beans will be ground with water until a slurry is formed and then filtration is done to separate the raw soya milk from the solid residue, i.e. okara. The raw soya milk required to be cooked for a few min before consumption. This step helps to prolong the shelf-life, improve the nutritional value, and improve the protein digestibility (Huang et al. 2006; Imran 2007; Toro-Funes et al. 2014; Peng et al. 2017). However, this traditional method of preparing soya milk leads to the development of beany flavour in the soya milk. This beany flavour, described as grassy, green, bitter, and astringent (Poliseli-Scopel et al. 2013), might not be acceptable to some consumers, especially those in western countries (Wang et al. 1998).

There are a few pre-requisite conditions required for the lipoxygenase (LOX) in soya bean to react and produce beany flavour. These include the release of LOX and polyunsaturated fatty acids from the soya bean, as well as the presence of water and oxygen (Imran 2007; Zheng and Brash 2010). LOX catalyses the oxidation of polyunsaturated fatty acids during soaking and grinding of the soya milk processing. The products from the oxidation are conjugated unsaturated fatty acid hydroperoxides and these can be easily degraded into volatile compounds, such as hexanol, trans-2-nonenal, 1-octen-3-ol, trans-2, trans-4-decadienal, 2-pentylfuran, 1-octen-3-one, and trans-2, trans-4-nonadienal, to cause beany flavour in the soya milk (Kumar et al. 2003; Yuan and Chang 2007; Zhang et al. 2012).

Several methods have been developed to inactivate LOX, including thermal treatments, such as hot grinding and blanching (Kwok and Niranjin 1995; Yuan et al. 2008; Zhang et al. 2012; Peng et al. 2017), non-thermal treatments, such as the use of high pressure (Wang et al. 2008) and irradiation (de Barros et al. 2014), genetic modification to produce soya beans with low or null LOX (Yuan and Chang 2007), and the use of masking agents. Among these methods, hot grinding is the most popular approach to produce low-beany soya milk (Lv et al. 2011; Zhang et al. 2012). However, the problem of beany flavour remains unresolved as the presence of water and oxygen during hot grinding allows some of the residue/remaining LOX to react with the polyunsaturated fatty acids, since it required time for the heat to completely inactive the LOX presence in soya bean (Baysal and Demirdöven 2007; Poliseli-Scopel et al. 2013).

Superheated steam transfers sensible heat to increase its temperature above the corresponding saturation temperature at a given pressure in order to dry product. Provided that the temperature is still higher than the saturation temperature at the operating pressure, a drop in temperature will not result in the condensation of steam. Moisture evaporated from the product becomes part of the medium and does not need to be removed (Pronyk et al. 2004). The drying medium of superheated steam is also free from oxygen (Van Deventer and Heijmans 2001; Mujumdar and Law 2010). This moisture- and oxygen-free environment is favourable since the absence of water and oxygen could prevent LOX from becoming active. As yet, no information is available on the use of superheated steam on soya beans to inactive LOX. Therefore, this study aims to optimise the superheated steam treatment in LOX inactivation of soya beans in order to produce low beany flavour soya milk.

Materials and methods

Chemicals and reagents

All chemicals used in this study were of analytical grade. All the chemicals were purchased from Sigma-Aldrich Co., St. Louis, USA.

Soya beans

Soya bean variety (Arius) grown in Quebec, Canada was supplied by Tan Ban Huat Sdn. Bhd., a soya bean wholesaler in George Town, Penang, Malaysia. This variety was selected for its high protein content (44%), which is well suited for soya milk production.

Experimental design and statistical analysis

Ranges of time (2–15 min) and temperature (100–140 °C) used in this study were adapted from Prachayawarakorn et al. (2006) with slight modifications. Software package, Design-Expert version 6.0.10 was used in this optimisation study. Face-centred central composite design (FCCD) with two variables and three levels were selected. The experiments consisted of 13 runs with two variables and five replicates of the central point for the pure error estimation (Table 1). Influence of factors on independent variables (heating time: X₁ and heating temperature: X₂) on the specific lipoxygenase (LOX) activities and crude protein content by superheated steam were described according to the equation below:

$$Y=b_0+b_1{X}_1+b_2{X}_2+b_{12}{X}_1{X}_2+b_{11}{\left(X_1\right)}^2+b_{22}{\left(X_2\right)}^2$$ 1

where Y is the dependent variable (response) predicted by the model, b₀ is the constant coefficient of intercept, b₁ and b₂ are the linear effects, b₁₁ and b₂₂ are the quadratic effects, b₁₂ is the linear-by-linear interaction, X₁ is the heating time (min), and X₂ is the heating temperature (°C).

Table 1.

Experimental design for the superheated steam treatment on the soya beans

Independent variables	Coded level (x)
	− 1	0	+ 1
X1: Time (min)	2.00	8.50	15.00
X2: Temperature (°C)	100	120	140

Run	Coded level x	Actual level (X1)	Coded level x	Actual level (X2)
1	− 1	2.00	− 1	100.00
2	0	8.50	0	120.00
3	0	8.50	0	120.00
4	0	8.50	+ 1	140.00
5	0	8.50	0	120.00
6	+ 1	15.00	+ 1	140.00
7	0	8.50	0	120.00
8	− 1	2.00	0	120.00
9	0	8.50	− 1	100.00
10	+ 1	15.00	− 1	100.00
11	+ 1	15.00	0	120.00
12	0	8.50	0	120.00
13	− 1	2.00	+ 1	140.00

Superheated steam treatment

Soya beans (170 g) were weighed and heated using superheated steam oven (Sharp, Healsio AX-1500 V, Japan) with different combination of heating time (2–15 min) and temperature (100–140 °C) by referring to the method used by Prachayawarakorn et al. (2006) with slight modifications. Prachayawarakorn et al. (2006) used superheated steam fluidised-bed system at temperatures of 120, 135, and 150 °C and drying times of 2, 5, 7, 10, and 15 min. Thirteen sets of heating time and temperature were generated by response surface methodology (RSM) software. By using the RSM, optimised superheated steam condition with lowest LOX activity in soya bean and highest crude protein content in soya milk was obtained and verified.

Determination of specific LOX activity

Specific LOX activity of raw soya beans and soya beans treated with superheated steam was determined using the method from Axelrod et al. (1981) and de Barros et al. (2014). Blended soya bean (10 mg) was macerated with 1 mL of 50 mM Tris–HCl buffer, pH 8.0 containing 20 mM CaCl₂. The macerated solution was then centrifuged at 5000g for 10 min at 4 °C, and the supernatant (enzymatic extract) was pipetted into an Eppendorf tube and stored in a freezer (Ardo, CV 382, Italy) at − 20 °C until further analysis. A stock solution of 10 M sodium linoleate was prepared by adding 156 μL linoleic acid and 180 μL Tween 20 into approximately 20 mL previously boiled deionised water. The solution was homogenised by repeated pipetting. Care was taken to avoid the bubbles formation. Drops of 0.5 N NaOH were added to clear the solution and then the solution was transferred to a 50-mL volumetric flask wrapped with aluminium foil and volume was topped up to 50 mL. The stock linoleate solution was then stored in Falcon tube wrapped with aluminium foil at − 20 °C (de Barros et al. 2014).

For specific LOX activity analysis, 100 μL of enzymatic extract (obtained from soya beans), 400 μL of the sodium linoleate stock solution (substrate), and 5 mL of 50 mM phosphate buffer (pH 6.5) were mixed in a test tube and transferred to a quartz cuvette for immediate analysis. The reaction rate of the specific LOX activity was determined using UV–Vis spectrophotometer (Shimadzu, UV-1650 PC, Japan) with an increase of absorbance at 234 nm over a period of 60 s at intervals of 20 s. The blank was consisting of only the same quantity of substrate and buffer and submitted to the same procedure (de Barros et al. 2014). The specific LOX activity was calculated by dividing the reaction rate obtained from UV–Vis spectrophotometry analysis by protein content.

The protein content of the enzymatic extract was quantified by the Bradford method, using bovine serum albumin (BSA) as a standard (Bradford 1976). Measurement of absorbance was carried out using UV-1650 PC UV–Vis spectrophotometer at wavelength of 595 nm. Relative measurement of protein concentration can be obtained by comparing with the BSA standard curve. The specific LOX activity was expressed as μmol/(min mg protein) and was calculated using the following formula:

$$\text{Specific}\text{activity}\text{of}\text{LOX}=\left(\mathrm{\Delta }A\times DF\times V_{\text{c}}\right)/\left(V_{\text{E}}\times L_{\text{C}}\times \epsilon \times C_{\text{P}}\right)$$ 2

where ∆A is variation in absorbance at 234 nm (1/min), V_E is volume of the enzymatic extract (L), DF is dilution factor, V_C is total volume in the cuvette (L), ε is the molar absorption coefficient of linoleic acid hydroperoxide at 234 nm, i.e. 0.025 L/(μmol cm), L_C is the path length of the cuvette, i.e. 1 cm, and C_P is the protein concentration (mg/L) quantified using Bradford method.

Soya milk preparation

Raw soya beans and soya beans treated with superheated steam (70 g) were soaked in water for overnight. Then it was blended using a blender (Panasonic, MX-355, Malaysia) with addition of 500 mL of water. After that, the soya milk was extracted using a muslin cloth. The raw soya milk was heated until boiled. The temperature of the soya milk was lowered down using cold water bath. The soya milk sample was stored at 4 °C until further analysis.

Determination of crude protein content in soya milk

The crude protein content of soya milk samples produced was determined using the Kjeldahl method with a conversion factor of 6.25 (AOAC Method 991.20). The percentage of crude protein (wet basis) was calculated using the formula below:

$$\%\text{Crude}\text{Protein}\left(\text{wet}\text{basis}\right)=\left(\left[V_{\text{S}}-V_{\text{B}}\right]\times M\times 14\times 100\times 6.25\right)/w$$ 3

where V_S and V_B are volume of HCl used for the sample and blank, respectively (L), M is the molarity of HCl (0.02 M), 6.25 is the nitrogen conversion factor, and w is the weight of sample (g).

Sensory evaluation

Directional Difference Test was used to evaluate which sample has a weaker beany flavour. Soya milk was served in preparation area temperature (~ 25 °C) with a volume of 50 mL in a small plastic cup coded with 3-digits random numbers (Wang et al. 1998). Two samples were used and presented at the same time in the sensory evaluation, whereby one was control sample (soya milk produced using non-superheated steam treated soybeans), while the other one was optimised sample (soya milk produced using soya beans treated under optimised superheated steam condition). Drinking water was used to rinse the mouth.

The sensory panellists comprised of 40 untrained panellists (students of the Food Technology Division, School of Industrial Technology, Universiti Sains Malaysia). The data obtained from the sensory evaluation was tabulated using Χ² (Chi-square) method by referring to the Table of Percentiles of the Χ² Distribution with (P < 0.05). The probability of choosing the correct and incorrect answer by chance is 0.5. The degree of freedom is 1, α = 0.05, and 1 − α = 0.950. The Χ² equation used is as below:

$${\text{X}}^2=\left({\left[{\text{O}}_1-{\text{E}}_1\right]}^2/{\text{E}}_1\right)+\left({\left[{\text{O}}_2-{\text{E}}_2\right]}^2/{\text{E}}_2\right)$$ 4

where Ο₁ is sum of the correct evaluation, Ο₂ is sum of the wrong evaluation, Ε₁ is sum of the assume correct evaluation, and Ε₂ is sum of the assume wrong evaluation.

Results and discussion

Model fitting

The effects of two process variables, i.e. heating time (X₁) and heating temperature (X₂) of superheated steam treatment on two responses, i.e. specific LOX activity of soya beans and crude protein content of soya milk, were studied during the experimentation (Table 2). From the results of analysis of variance (ANOVA) obtained for the LOX activity (Table 3), P value for the model was significant (P < 0.05) and for lack of fit was 0.4648, which was not significant (P > 0.05). These indicated that the model fitted the data well and accurate in predicting the relevant responses. The R² value for this response variable was 0.9860, which is greater than 0.80 and showed that the model and the data obtained were well-fitted. This indicated that the reaction could be explained by the regression model. The adjusted and predicted R² value were 0.9759 and 0.9386, respectively. Hence, the model is good in predicting the response as the difference between the adjusted R² and predicted R² was less than 0.20. The CV of the LOX activity in this experiment was 5.80%, which is less than 10%. This determined that the reliability of this response in this experiment was very high. The fitted quadratic model for specific LOX activity, expressed as μmol/(min mg protein), in soya beans in actual values is given in the following equation:

$$\text{Specific}\text{LOX}\text{activity}=0.147-1.187\text{E}-3X_1-1.968\text{E}-3X_2-2.692\text{E}-6X_1{X}_2+4.783\text{E}-5{\left(X_1\right)}^2+7.552\text{E}-6{\left(X_2\right)}^2$$

where X₁ is the heating time (min) and X₂ is the heating temperature (°C).

Table 2.

Experimental value for the specific lipoxygenase (LOX) activity of the soya bean and crude protein of the soya milk

Independent variables	Coded level
	− 1	0	+ 1
X1: Time (min)	2.00	8.50	15.00
X2: Temperature (°C)	100	120	140

Run	Coded level		Specific LOX activity [μmol/(min mg protein)]	Crude protein (%)
	x	x
1	− 1	− 1	0.0226	3.5
2	0	0	0.0094	3.2
3	0	0	0.0098	3.2
4	0	+ 1	0.0096	2.8
5	0	0	0.0093	3.2
6	+ 1	+ 1	0.0067	1.9
7	0	0	0.0109	3.3
8	− 1	0	0.0170	3.4
9	0	− 1	0.0176	3.4
10	+ 1	− 1	0.0141	3.3
11	+ 1	0	0.0082	2.7
12	0	0	0.0106	3.2
13	− 1	+ 1	0.0166	3.3

Table 3.

ANOVA for response surface quadratic model on the specific lipoxygenase (LOX) activity of soya beans and crude protein content in soya milk by superheated steam treatment

Source	df	Sum of square	Mean square	F value	P value Prob. > F
Specific LOX activity of soya beans
Model	5	2.58 × 10−4	5.16 × 10−5	98.36	< 0.0001 significant
Residual	7	3.67 × 10−6	5.24 × 10−7
Cor total	12	2.62 × 10−4
Lack of fit	3	1.16 × 10−6	5.37 × 10−7	1.04	0.4648 not significant
Pure error	4	2.06 × 10−6	5.15 × 10−7
R2	0.9860
Adjusted R2	0.9759
Predicted R2	0.9386
C.V.	5.80
Adequate precision	32.934
Crude protein content in soya milk
Model	5	2.160	0.43	107.88	< 0.0001 significant
Residual	7	0.028	4.01 × 10−3
Cor total	12	2.190
Lack of fit	3	0.020	6.68 × 10−3	3.34	0.1372 not significant
Pure error	4	8.00 × 10−3	2.00 × 10−3
R2	0.9872
Adjusted R2	0.9780
Predicted R2	0.9066
C.V.	2.04
Adequate precision	36.112

According to the results of ANOVA for the crude protein content in soya milk (Table 3), the P value for the model obtained was significant (P < 0.05) and the P value for the lack of fit was not significant (P > 0.05) which was 0.1372. It showed that the model was precise in predicting the relevant response. The R² value obtained for crude protein was greater than 0.80, which was 0.9872. This further strengthens the actual data matched the model. Besides, the adjusted and predicted R² were 0.9780 and 0.9066, respectively. The adjusted R² was greater than 0.90 while the difference between the adjusted R² and predicted R² was less than 0.20, this further strengthen the relationship between the model and the response values. The CV of the crude protein in soya milk by superheated steam was only 2.04%. This indicated that there is a high reliability in this experiment for the response of crude protein content in the soya milk. The fitted quadratic model for crude protein content, expressed as  % wet basis, in soya milk in actual values is given in the following equation:

$$\text{Crude}\text{protein}\text{content}=0.221+0.273X_1+5.30\text{E}-2X_2-2.308\text{E}-3X_1{X}_2-3.224\text{E}-3{\left(X_1\right)}^2-2.155\text{E}-4{\left(X_2\right)}^2$$

where X₁ is the heating time (min) and X₂ is the heating temperature (°C).

Analysis of response surfaces

The plot of the comparison of the predicted value and the experimental value can be used to indicate the closeness between the value and the accuracy of the model used. Figure 1a, b show the fitted-line plot of predicted versus accuracy for specific LOX activity of soya beans and crude protein content of soya milk, respectively, in this study. The plots showed that the closeness between the predicted values and the experimental values proved that the response surface models in this experiment were adequate in predicting both the specific LOX activity and crude protein content by superheated steam treatment.

Fig. 1.

Fitted-line plot indicating the closeness between predicted values and experimental values for the a specific lipoxygenase (LOX) activity and b crude protein content by superheated steam treatment

The effects of the independent variables; A: heating time and B: heating temperature of the superheated steam treatment, on the specific LOX activity and crude protein content were illustrated in Fig. 2a, b, respectively. The specific LOX activity was seen to be decreasing with the increasing time and temperature of the superheated steam treatment (Fig. 2a). It was obvious that the specific LOX activity was the highest when the time was short and the temperature was low, and the lowest obtained when the time was long and temperature was high. As mentioned earlier, LOX is a group of enzymes responsible to produce beany flavour in soya milk (Kumar et al. 2003). Thus, LOX is sensitive to heat and could be inactivated at higher temperature (Wang et al. 2008; Lv et al. 2011). According to Imran (2007), a treatment with temperature around 100 °C for 5–10 min is effective in inactivating the LOX in soya bean. These findings showed that the superheated steam treatment could reduce the specific LOX activity as LOX is a heat-sensitive enzyme.

Fig. 2.

Three-dimensional plots for a specific lipoxygenase (LOX) activity of soya beans and b crude protein content of soya milk as a function of heating time and heating temperature of the superheated steam treatment

The response of the crude protein towards the independent variables, A: heating time and B: heating temperature, was quite stable when the time was short and temperature was low (Fig. 2b). When the time and temperature of the superheated steam treatment were lower than 2 min and 100 °C, the effect on crude protein content seemed to be negligible. The crude protein content in the soya milk depends on the protein solubility in the soya beans. When the protein solubility in the soya bean is low, it would lead to a low crude protein content in soya milk as most of the protein remains insoluble in the okara. The crude protein solubility decreased faster when the heating time and temperature increased (Fig. 2b). One possible explanation to this finding is the destruction of the helical regions of the protein and the subsequent disulphide bonds between the amino acid groups (Prachayawarakorn et al. 2006).

Validation of predictive model

The optimum superheated steam conditions provided by the respond surface methodology (RSM) were as follow: heating time at 9.3 min and heating temperature at 119 °C. Considering the operating convenience for superheated steam oven used in this study, the optimal values of variables were adjusted as follow: time at 9.5 min and temperature at 120 °C. To validate the predicted model, tests were carried out at the suggested optimal conditions for time and temperature. The predicted specific LOX activity and crude protein content were 0.0098 μmol/(min mg protein) and 3.2%, respectively (Table 4). The specific LOX activity was verified at 0.0081 ± 0.0002 μmol/(min mg protein) with a total reduction of 66% from the control (i.e. raw soybean), which was 0.0236 μmol/(min mg protein). On the other hand, the crude protein content was verified at 3.0 ± 0.1% with a total reduction of 23% from the control (3.9%).

Table 4.

Predicted and experimental values for the specific lipoxygenase (LOX) activity and crude protein content

	Condition		Response
	Time (min)	Temperature (°C)	Specific LOX activity [μmol/(min mg protein)]	Crude protein content (%)
Predicted	9.3	119	0.0098	3.2
Experimental*	9.5	120	0.0081 ± 0.0002	3.0 ± 0.1

*Data written as mean ± standard deviation (n = 3)

By comparing the predicted and experimental values for both the specific LOX activity and crude protein content, the experimental values are slightly lowered than the predicted ones. This could be due to the adjustment on the heating time and temperature due to the limitation of the superheated steam oven, whereby the time and temperature were slightly longer (i.e. 9.5 min) and higher (i.e. 120 °C), respectively, than the predicted optimal superheated steam conditions (i.e. 9.3 min for 119 °C).

Sensory evaluation

The sensory evaluation was carried out using Directional Different Test with only 2 samples; control and optimised samples. The optimised sample was produced from the soya bean treated with the optimum superheated steam condition generated using RSM with the setting of minimum specific LOX activity in the soya bean and maximum protein content in the soya milk (i.e. heating time = 9.5 min and heating temperature = 120 °C). As LOX is the main component that is responsible for the beany flavour in soya milk (Poliseli-Scopel et al. 2013), lowering the LOX activity could help in reducing the beany flavour of the soya milk. Thus, this sensory test was to determine whether the optimised sample of soya milk has a weaker beany flavour than the controlled sample of soya milk.

Out of the 40 panellists, there were 28 correct responses. The data of the sensory evaluation was tabulated using Χ² (Chi-square) Method. By referring to the Table of Percentiles of the x² Distribution at probability 0.05, the Χ² from the calculation was higher than the result obtained from the Table of Percentile of the Χ² Distribution (Table 5). Thus, there was significant difference (P < 0.05) in term of beany flavour between the two soya milk samples. This result showed that the optimised soya milk sample has a significant (P < 0.05) weaker beany flavour than the controlled sample.

Table 5.

Criteria and results from Χ² (Chi-square) calculation

Criteria	Results
v	1
Probability	P < 0.05
α	0.05
1 − α	0.95
Χ2 from the table of percentiles of the Χ2 distribution	3.84
Χ2 from the calculation	6.40

From the results of the sensory evaluation, it showed that the optimised superheated steam treatment could reduce the activity of the LOX during the production of soya milk. This might due to the denaturation of the LOX structure during superheated steam treatment and this caused the enzyme to be inactivated (Zhang et al. 2012; Peng et al. 2017). By inactivating the enzyme LOX, it would not be able to react with its substrate such as linoleic and linolenic acids during the soaking, where both water and oxygen are available for the reactions (Zheng and Brash 2010). Therefore, lesser amount of volatile beany flavour compounds was produced during the production of optimised soya milk as the amount of active LOX had been reduced. Hence, the beany flavour of the optimised soya milk was significantly (P < 0.05) weaker than the controlled soya milk sample.

Conclusion

Optimised superheated steam condition for treating soya bean was obtained at 119 °C for 9.3 min with the lowest specific LOX activity and highest crude protein content. Due to the limitation on the setting of the superheated steam oven, the optimum condition was set to 120 °C for 9.5 min and the results. The crude protein content of the soya milk produced under the optimised condition was 3.0 ± 0.1%, fulfilling the standard as stated in Regulation 357 of Malaysian Food Regulations 1985. With the reduction in the specific LOX activity from 0.0236 to 0.0081 μmol/(min mg protein), a significant weaker (P < 0.05) beany flavour soya milk was produced.