Use of D‐optimal combined design methodology to describe the effect of extraction parameters on the production of quinoa–barley malt extract by superheated water extraction

Abstract Superheated water extraction was applied to produce quinoa–barley malt extract. D‐optimal combined design was used to optimize the extraction conditions (time (min), solid–water ratio and particle size to obtain maximum protein and carbohydrate content, and minimum turbidity and pH. Quinoa flour (10%–30%), barley malt flour (70%–90%), different particle sizes (F = 420 µm, G = 710 µm), time (15–45 min), and solid–water ratio (0.1–0.2) were selected as independent variable and protein, carbohydrate, turbidity, and pH as dependent factors. Polynomials models satisfactorily fitted the experimental data with the R 2 values of .9961, .9909, .9949, and .9987, respectively. The protein and carbohydrate value was affected by superheated water extraction parameters. Our results revealed that increasing quinoa/barley malt ratio has significant effect on the turbidity and pH. The optimum extraction conditions were quinoa flour (30%), barley malt flour (70%), solid–water ratio (0.2), time (45 min), and particle size (F = 420 µm).

Barley ranks fourth in quantity produced and cultivation area of cereals in the world and has a long history of use as a source of human nutrition. Clinical studies have shown that the intake of β-glucan from either barley flour or barley malt products can control cardiovascular disease (CVD), type 2 diabetes, obesity, colorectal cancer, and lower blood glucose levels in humans (Arendt & Zannini, 2013;Chappell et al., 2017;De Arcangelis et al., 2019;Suriano et al., 2018).
There are different methods for extracting the functional compounds from cereal or pseudocereal, but most of them were traditionally extracted by organic solvents such as methanol, ethanol, and acetone. However, conventional methods are time-consuming and some of the solvents are toxic. Recently, advanced extraction technologies such as ultrasound, supercritical CO2, microwave, and superheated water extraction (SWE) have been developed to overcome these problems.
Subcritical water extraction or superheated water extraction (SWE) has become a popular green extraction technique for the isolation of different classes of compounds from natural matrices. Low price, safety and green character of water, good yields of target compounds, and reduced energy consumption make this technique favorable for potential industrial applications (Nastić et al., 2018). SWE technique has been studied in recent years to extract functional compounds from diverse matrix of food by-products and agricultural products (Erşan et al., 2018;Miró-Abella et al., 2018), such as bagasse waste, potato peel, wheat straw, tea industry by-products (Naffati et al., 2017), ginger extract (Švarc-Gajić et al., 2017), soybean products (Moras et al., 2017), traditional Serbian medicinal plants (Nastić et al., 2018), and Pacific oyster (Getachew et al., 2019).
There have been statistical and soft-computing approaches to find out the multidimensional correlation functions. Mixture design is one of the most important statistical techniques used to optimize the ingredients found in the formula considering alternative ingredients. For this, mixture design and RSM can be combined to determine the effects of extraction processing factors. So, combined design was used for obtaining many detailed information about relationship between many independent and dependent factors at specified conditions Ghelichkhani et al., 2019;Icyer et al., 2016;Rafiei Nazari et al., 2018).
There have been a few reports on the extraction of functional substances from quinoa and barley malt by using superheated water extraction for production of functional beverage. The objective of this study was to support commercial application of quinoa to produce a new functional beverage and to evaluate the potential of SWE as environmentally friendly method for the production of quinoa-barley malt extract. Furthermore, the present study aimed at investigating the effect of SWE parameters including extraction time, particle size, and solid-liquid ratio on yield extraction, and optimized these condition by using D-optimal combined design.

| Materials
Quinoa seeds of Sajama variety were obtained from The Karaj Seed and Plant Improvement Institute, Karaj, Iran, in September 2018.
Barley malt was purchased from Beh malt Co., and all chemicals used were of reagent grade obtained from Sigma and Merck.

| Sample preparation and quinoa/barley malt extraction
Quinoa seeds were separated from impurities (leaves, stones, etc.) and soaked in water for 24 hr and washed thoroughly to remove saponins (foamless) and then dried in ambient temperature. After that, quinoa seeds were ground into powder by an electric grinder (IKA 1603600M 20 Universal Mill, 230V). The milled quinoa flour and barley malt flour were defatted with hexane as solvent in a ratio of 1:5 among 24 hr with the aid of a fattened shaker (Fisher Scientific Ltd,. Afterward, the fat-free flours were placed for 24 hr in an oven at 40°C, to isolate the residues of solvent, and to obtain good powdered flours. Then all the flours were sieved, using different mesh size sieves (mesh N° (ASTM E11) 40, sieve size 420 µm and No. (ASTM E11) 25, sieve size 710 µm, Sarv Azma Co).
Each sample was stored at −18°C until use. Superheated water extraction was performed using Synth wave apparatus (Milestone) with some modification. Quinoa/barley malt were mixed with tap water, the mixture was stirred by an industrial blender at room temperature subtracting the percentage sum of moisture, protein, fat, crude fiber, and ash from 100%. Turbidity of samples was measured using a turbidometer (2100N Turbidimeter, HACH, CO) and reported in terms of nephalometric turbidity units (NTU). All the analyses were done in triplicate.
Samples were analyzed for total starch using the microwaveassisted sonication/iodometric USDA Research method (ICUMSA GS1-16) (Eggleston & Triplett, 2017). where Y1 (protein), Y2 (carbohydrate), Y3 (turbidity), and Y4 (pH) are the predicted response; i is a linear coefficient, and ij is a quadratic coefficient. i X i represents the linear blending portion, and the parameter ij X i X j represents the excess response over the linear model due to the interaction between two components, and this effect is often called antagonism ( ij < 0) or synergism ( ij > 0) (Moreira et al., 2007;Santafé-Moros et al., 2005).

| Experimental design and statistical analysis
RSM was used for modeling of the processing factors. In this method, the statistical data indicating the correlation between independent factors and the response were adjusted to fit the second-order polynomial equation in Equation (2) for RSM: in Equation (2); i and j are the linear quadratic coefficient; X i and X j are the encoded independent variables; k is the number of studied factors optimized; β 0 is the constant (intercept); β i the linear coefficient; β j , jj , and ij are interaction coefficient of linear, quadratic, and second-order terms, respectively; X i and X j are independent variables and ε is the error. Statistical significance of the model parameters was set at the 5% (p-value < .05).

| Goodness of fit models
Statistical analysis parameters of the Fisher test value (F-value), pvalue of model, coefficient of determination (R 2 ), the coefficients of determination (R 2 -adj), predicted R 2 , p-value of lack of fit(LOF), adequate precision, and CV% obtained from the analysis of variance (ANOVA) were used for evaluation of the goodness of fit in models.
Fitted polynomials models were applied for optimization of extraction conditions. Statistical significance of the model parameters was set at the 5% (p-value < .05). Table 2 reports the corresponding 37run D-optimal combined experimental plan and obtained responses.
Protein content response was fitted to Quadratic x Quadratic model, which mean that both the mixture and process factor fit to quadratic models. According to Table 3, Table 5. For this response, A, B, AC, AD, BC, BD, AC 2 , and BC 2 are significant model terms.
Turbidity was fitted to Linear x Quadratic model. According to the ANOVA analysis shown in Table 3, the model is significant as model p-value less than a 0.05. In this case, A, B, AC, AD, BC, BD, CD, ACD, ACE, ADE, BCE, BDE, AC 2 , AD 2 , BC 2 , and BD 2 are significant model terms.
( The quality of fitting to the selected models was confirmed by the attained high values for the coefficient of determination obtained for protein content (R 2 = .9961), carbohydrate content (R 2 = .9909), turbidity (R 2 = .9949), and pH (R 2 = .9987) are presented in Table 3.

F-value of model
High R 2 value shows higher correlation between the independent variables and the responses.
Adjusted coefficient (Adj-R 2 ) was for protein content (.9916), carbohydrate content (.9828), turbidity (.9903) and pH (.9975). The value of the adjusted coefficient of determination (R 2 Adj) also indicates an excellent correlation, supporting that this model explains the experimental results adequately. Table 3, all of the predicted R 2 of responses were above of .80. High values of predicted R 2 represent a high degree of correlation between the experimental and predicted values.

As shown in
According to Table 3, for all responses p-values of LOF were higher than .05. Non-significant lack of fit of responses is good for models. Adequate precision measures the signal-to-noise ratio. A ratio greater than 4 is desirable (Wang et al., 2018). According to Table 3, adequate precision of the responses ranged from 36.00 to 80.56. These results imply the validity of models.
As seen from The predicted models for the protein content (Y1), carbohydrate content (Y2), turbidity (Y3), and pH (Y4) of the extracts are presented in Table 4    shows that the minimum amount of protein was produced when this ratio is 10/90. As tabulated in Table 5, the protein content of quinoa flour (15.76 g/100 g dry basis, d.b.) was higher than barley malt (11.13 g/100 g d.b.). Generally, quinoa seeds have a higher nutritional value, in comparison with most cereals. The quinoa protein content varies from 14 to 20 (g/100 g d.b.) (Matiacevich et al., 2006) which on average is higher than common cereals such as rice, wheat, and barley malt (Valencia-Chamorro, 2003). As shown in Figure  The test results as reflected in Figure 2 show the interaction between solid-water ratio and time (CD) on the protein content while the other independent variable was constant at 20% quinoa, 80%

| Response 1, protein content
barley malt, and average of particle size F and G. From Figure 2, it F I G U R E 2 The effect of time and solid-water on protein (%)

F I G U R E 3
The effect of quinoa/barley malt flour on carbohydrate (%) (both particle size F and G)

F I G U R E 4
The effect of time and solid-water on carbohydrate (%) (both particle size F and G)

F I G U R E 1 The effect of quinoa/barley malt flour on protein (%)
can be concluded that the percentage of protein increased with increasing extraction time and solid-water ratio as processing factor.
Maximum protein was extracted when both factors were at high level (Time = 45min and solid-water ratio = 0.2). It means both factors have a positive effect on the protein content.  Under hydrothermal conditions such as subcritical water extraction method, carbohydrates undergo rapid hydrolysis to monosaccharides, which are then further degraded. Starch and hemicelluloses are hydrolyzed much faster than cellulose, which in contrast to the former two has a mainly crystalline structure (Toor et al., 2011). The degradation of carbohydrates to glucose and other saccharides in sub-and supercritical water has been reviewed by several authors (Behrendt et al., 2008;Bröll et al., 1999;Yu et al., 2008). Figure 4 shows the effect of extraction time (C) and solid-water ratio (D) on this response at 20% quinoa flour, 80% barley malt, and average of particle size F and G. According to this figure, when both the solid-water ratio and extraction time increase, the carbohydrate content increased. Maximum carbohydrate yielded when solidwater ratio was at high level (0.2) and time was at the highest level (45 min).

| Response 3, Turbidity
The effect of the quinoa/barley malt ratio on turbidity (NTU) at extraction time 30 min, solid-water ratio 0.15, and particle size G is depicted in Figure 5. It can be observed that an increase in the quinoa content and decrease in barley malt in formulation of extract linearly increases the turbidity (NTU) of sample, the minimum amount of turbidity (NTU) is obtained when quinoa/barley malt ratio is 10/90, and turbidity is at maximum when this ratio is 30/70.
Protein is one of the important factor that affects turbidity in beverages, proteins have been associated with hazes in beer, red and white wine, apple, grape juice, and kiwifruit juice (Siebert & Lynn, 2000). According to Table 4 Figure 7 illustrates the effect of quinoa/barley malt ratio on pH response at 20% quinoa, 80% barley malt in average of particle size F and G. The test results indicate that by increasing the quinoa/barley malt ratio, pH increased. According to Table 5, the pH of quinoa is 6.40 and higher than the pH of barley malt that is 6.07. Results presented in Figure 7 show that the minimum of this response is obtained when quinoa/barley malt ratio is 10/90 and the maximum of pH was produced when this ratio was 30/70.

| Response 4, pH content
The response surface is plotted to study the interaction between concentration and time (CD) on the pH (Figure 8). The test results indicated that by increasing the extraction time pH decreased, by increasing the solid-water ratio pH increased. The minimum pH was obtained when extraction time and solid-water ratio are at high level (time = 45 min and solid-water ratio = 0.2). Decomposition of biomass, carbohydrates, and amino acids under subcritical water condition produces acidic component such as organic acid. Due to the formation of water-soluble organic acids, the pH of the treatment medium may be decreased (Lamoolphak et al., 2006;Pourali et al., 2009;Salak Asghari & Yoshida, 2006).

| OP TIMIZ ATI ON OF E X TR AC TI ON CONDITIONS
The main purpose of this study was to determine the optimum extraction conditions time (min), solid-water ratio and particle size of quinoa-barley malt flours. Derringer's desirability functions were applied for the selection of optimum extraction conditions for production of fermented probiotic beverages in next step. Overall Derringer's desirability function was established for the criteria that maximum protein and carbohydrate content and minimum turbidity and pH (Granato & Ares, 2014 In order to validate the results of optimization, three replicates of experiments for each response were performed under the optimal conditions. The comparison of experimental results with the predicted results under optimal condition is carried out by Paired-t test. The results of validity evaluation of optimal conditions are summarized in Table 6. As shown, insignificant differences between experimental results and predicted results imply on validity of optimal conditions.

| CON CLUS IONS
This study was aimed to optimize the production of quinoa-barley malt extract by using superheated water extraction method. The designed model results proved that D-optimal combined design methodology is a very effective method to reach very accurate results in the design of superheated water extraction condition.
Results from responses (protein, carbohydrate, turbidity, and pH) demonstrated the regression models were accurate within the set of values of the included variables. The result indicated that increase in quinoa/barley malt ratio increase the protein content whereas the carbohydrate values linearly decrease. Quinoa/barley malt ratio also affected on physicochemical properties such as turbidity and pH. Both of these parameters increased when quinoa/ barley malt ratio increased. In this study, the protein and carbohydrate value was affected by superheated water extraction parameters. By increasing solid/water ratio and extraction time, protein and carbohydrate content increased. The optimal SWE conditions at quinoa 30%, barley malt 70%, solid-water ratio 0.2% at 45 min, and particle size F are obtained. It is necessary to select the suitable extraction technique and optimize the extraction parameters in order to obtain the highest yield of valuable material. In conclusion, although SWE appears to be a useful extraction method for production of valuable materials from cereal and pseudocereal, this method has only been used at pre-commercial scale and more investigation is required to study the quantity, quality, and stability of the extracted valuable materials to scale it up for industrial means.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are openly available in a public repository that does not issue DOIs, as mentioned in Table 2.