Applying Mechanistic Models in Bioprocess Development - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 157

Fig. 7 Regression correlation coefficient (R 2 ) for each model output, indicating the goodness of
the linear regression used for estimating the sensitivity of each model output to various
parameters. For R 2 values lower than 0.7, the corresponding standardized regression coefficient
(SRC) may yield erroneous information
for ethanol falls under the threshold, similarly to what was observed for glucose at
its depletion.
In Fig. 8 , an overview of the SRCs for each parameter and model output is
presented. Interpretation of parameter ranking and SRC should be made cau-
tiously. All model outputs seem to be sensitive to the yield coefficient of biomass
on oxidized glucose, even during the growth phase on ethanol (after glucose
depletion).
The ranking of each parameter according to the SRC for each model output is
illustrated in Fig. 9 . When analyzing this ranking, it is possible to see the decrease
in sensitivity of the glucose prediction towards the maximum glucose uptake rate,
as well as the simultaneous increase in sensitivity towards the maximum oxygen
uptake rate, during the growth phase on glucose. This is in agreement with the fact
that the consumption of glucose is initially only limited by the maximum uptake
rate (excess of glucose in the media), and afterwards as the biomass concentration
increases and glucose concentration decreases, the observed uptake rate is no
longer maximal. Similar figures for the parameter ranking regarding ethanol,
oxygen and biomass can be drawn.
With regard to the model predictions for ethanol, this model output is most
sensitive to the maximum glucose uptake rate and biomass yield on glucose
(reduction pathway) during the first growth phase, and later on the maximum
ethanol uptake rate. This is in good agreement with the fact that the production of
ethanol is a result of the reduction of glucose, and its consumption only takes place