3.4. Screening and optimization of the selected regulators
The 11 selected regulators were evaluated using a Plackett-Burman design
for their effects on the yield of ethanol mol/mol carbon (Table S9).
Table S9 demonstrates that the design matrix is chosen to screen
significant variables for ethanol production and the corresponding
response. The model appropriateness was assessed, and the variables
indicating statistically significant impacts were screened via their
p-values of ANOVA. Factors with confidence levels above 85% (P
< 0.15) were considered to have significant effects and were
therefore selected for further optimization studies. Imidazole was
estimated to be the most significant factor, with a probability value of
0.0008, followed by glycine (0.04), and glycerol (0.13). The other
insignificant variables were dismissed, and 2-level full factorial
design then calculated the optimum values for the three variables. We
evaluated the eleven best-anticipated regulatory interactions, including
two regulators with a strong in vitro impact in S. cerevisiaewith no prior evidence for their regulatory significance. The majority
of these regulators had not been already proposed in S.
cerevisiae .
To identify the right combination of regulators for optimal ethanol
production, the concentration of the aforementioned variables has been
optimized. In this study, a 23 two-level full
factorial design was used for three independent variables at two levels
each, resulting in 8 experiments. Table S4 presents the full
experimental plan regarding their actual and coded values, and the
corresponding results of experiments. ANOVA statistical results for the
selected factorial model are given in Table 1. The Model F-value of
697.8 indicated that the model was significant, so there was only a
0.014% probability that a ’ Model F-value ’ could occur due to noise,
according to the statistical results for the quadratic models provided
in Table 1. The values of ’prob > F’ for a model below 0.05
(< 0.0001) suggested that the model was statistically
significant, with a confidence interval of 99.99%.
Significant terms are shown in Table 1 indicates that interaction terms
of BC and AB are very significant, respectively. In Table S10, the
optimal concentrations suggested by the statistical model are provided,
and it was expected that the optimal yield of 0.36 mol/mol carbon was
obtained, which was verified experimentally. Additionally, ethanol
concentrations improved from approximately 5.5 g/l to 11.6 g/l. The
findings show that the approach was able to adjust the flux distribution
at optimum pH by adding certain regulatory compounds so that more
ethanol was produced, and 2.1 fold higher than the measured yield for
the control sample was achieved.