3.2. Optimization of the Enzymatic Hydrolysis Process
A Box–Behnken design (BBD) was used to analyze the experimental data. BBD belongs to Response Surface Methodology (RSM), which is a collection of statistical and mathematical techniques useful for process development, improvement, and optimization [
28]. In general, RSM has two main types of designs—the Box–Behnken design (BBD) and the central composite design (CCD). BBDs differ from CCDs in that they use fewer series and only three levels, compared to CCD’s five [
29]. For this reason, BBD was preferred in the present study.
All experiments were repeated three times and the means and standard deviations were calculated. Experimental data were evaluated statistically using Design-Expert software (version 13.0.5.0).
Table 3 presents the results of the studies on the radical scavenging activity of the obtained hydrolysates, against DPPH and ABTS.
DPPH and ABTS tests are some of the most widely used methods to assess antioxidant activity. The radical scavenging activity to DPPH in the existing LPH sample varied from 49.48 to 77.08%. Under the conditions of the analysis, the concentration of the standard antioxidant Trolox, at which 50% inhibition of the DPPH radical is reached, is 13.07 µg/mL. When analyzed with ABTS, the RSA% values of the hydrolysates ranged from 20.49 to 42.54%. Trolox inhibited 50% of the ABTS radical at a concentration of 64.70 µg/mL. As can be seen from
Table 3, all tested LPH samples inhibited both radicals, but to a different extent depending on the applied enzymatic hydrolysis parameters.
Table 4 summarizes the models obtained in this study. These models were evaluated based on the values of a lack of fit and coefficient of determination (R
2). The significance of each coefficient in the model was determined using an F-test obtained from the analysis of variance (ANOVA) generated. Numerical optimization was conducted to obtain the optimal conditions for the enzymatic hydrolysis process.
The software product Design-Expert defines the dependence of DPPH on the input variables as quadratic (
Table 5), with statistically insignificant coefficients in front of the added members. The Model F-value of 5.30 implies the model is significant. There is only a 0.60% chance that an F-value this large could occur due to noise.
P-values less than 0.0500 indicate model terms are significant. In this case, A, AB, BC, and B
2 are significant model terms.
The fit statistics for the obtained model are presented in
Table 6. The values of R2 and adjusted R
2 indicate a good explanation of the variability by the selected model for DPPH. The Adeq Precision value measures the signal-to-noise ratio and it must be greater than 4. The current ratio of 7.993 indicates an adequate signal. Therefore, this model can be used to navigate the design space.
The diagnosis of the model is realized through three types of graphs—a normal probability plot, a plot of the residuals versus the ascending predicted response values, and a graph of the predicted response values versus the actual response values—which are presented in
Figure 1. The normal probability diagram (
Figure 1a) shows whether the residuals follow a normal distribution. In this case, they follow a straight line with minor deviations, which shows that the residuals have a normal distribution. The residuals vs. predicted plot (
Figure 1b) shows that the residuals are randomly located around the line residual = 0. This suggests that the resulting relationship model is reasonable. Also, no residue “stands out” from the underlying random pattern of residues. This assumes there are no outliers. Finally, the actual vs. predicted plot (
Figure 1c) shows that the points are not quite close to the diagonal line, but are still within reasonable limits. This is understandable since R
2 = 0.8125.
In
Figure 2, a series of model graphs is presented. The perturbation plot (
Figure 2a) shows the influence of the input factors on the corresponding response. In this case, on the DPPH, the most significant influence is factor A (papain concentration), followed by factor B (temperature). The factor C line is almost parallel to the
x-axis and therefore has no significant influence.
The response surface plots (
Figure 2b–d) visualize the variation of the values of two independent variables within the experimental domain while holding the other one constant.
Figure 2b reveals that the maximum value of DPPH can be achieved by keeping the temperature and papain concentration at 40 °C and 2.0%, respectively, while the hydrolysis time is chosen to be a minimum of 60 min. It appears that Factor C has the opposite effect on DPPH. Also shown with flags are the actual and model-predicted values at two experimental points, demonstrating the minimum prediction error.
Figure 2c shows that the maximum value of DPPH can be achieved with a minimum hydrolysis time, papain concentration in the range of 1–2.5%, and maximum temperature.
Figure 2d reveals two regions with maximum DPPH values. Getting into one (bottom right) requires a hydrolysis time of over 90 min, a temperature of 20 °C degrees or less (which is outside the considered range), and a maximum papain concentration. To get into the other area (located on the opposite diagonal), a minimum time for hydrolysis (60 min), a temperature above 35 °C, and a minimum concentration of papain are required.
Two-factor interaction terms have been obtained to describe the dependence of ABTS on the input variables (
Table 7). The Model F-value of 16.95 implies the model is significant. There is only a 0.01% chance that an F-value this large could occur due to noise. This model includes the A, B, C, and BC terms, which are significant.
The fit statistics for the obtained model are presented in
Table 8. The values of R
2 and adjusted R
2 indicate a good explanation of the variability by the selected model for ABTS. The Adeq Precision value of 15.8639 indicates an adequate signal. Therefore, this model can be used to navigate the design space.
The diagnosis of the model is shown in
Figure 3. The normal probability plot (
Figure 3a) shows that the residuals have a normal distribution because they follow a straight line with small deviations. The plot of the residuals against the predicted (
Figure 3b) shows that the residuals are randomly located around the line residual = 0. This suggests that the resulting relationship pattern is reasonable. Also, no residues “stand out” from the random residue pattern. This assumes there are no outliers. Finally, the actual versus predicted plot (
Figure 3c) shows that the points are close to the diagonal line.
In
Figure 4, a series of model graphs is presented. The perturbation plot (
Figure 4a) shows that all three factors have approximately the same influence on the response ABTS, as they have almost the same slope.
Figure 4b reveals that the maximum value of ABTS can be achieved by keeping the temperature and papain concentration at 40 °C and 1.0%, respectively, while the hydrolysis time is chosen to be a minimum of 60 min.
Figure 4c shows that the maximum value of ABTS can be achieved with a maximum hydrolysis time (180 min), papain concentration in the range of 1–1.5%, and minimum temperature. From
Figure 4d, it can be seen that the maximum ABTS value is obtained in the following conditions: the minimum temperature value (20 °C), maximum hydrolysis time (180 min), and 1.0% papain concentration, or at the maximum temperature value (40 °C), minimum hydrolysis time (60 min), and 1.0% papain concentration.
A numerical optimization, which finds a point that maximizes the desirability function, was made to find the optimal process conditions.
Table 9 presents the specific optimum conditions for obtaining the highest level of bioactive compounds. The Design Expert software returns a table of 100 possible solutions. For brevity, only the first ten of them are presented here.