Applying Mechanistic Models in Bioprocess Development - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 159

Fig. 9 Ranking of each model parameter according to the magnitude of the SRC for each model
output: a rank of 1 indicates that the model output is most sensitive to that parameter, while a rank
of 11 indicates that the parameter contributes the least to the variance of the model output
during the second growth phase following the depletion of glucose. A similar
pattern was observed with regard to the model predictions for oxygen.
To analyze the sensitivity of the outputs to the parameters in more detail, two
time points during the exponential growth phase on glucose (t = 17 h) and on
ethanol (t = 27 h) were selected. The SRC and corresponding rank position for
these time points are provided in Table 9a and b , respectively. As could be
expected, during the growth on glucose, the parameters that most influence the
prediction of glucose are the biomass yield parameters (for the two pathways) and
the maximum uptake rate. The two yield coefficients have, however, a different
effect on the glucose prediction: while an increase in the oxidative yield will lead
to a lower predicted concentration, an increase in the reductive yield seems to
imply an increase in the predicted concentration. This may reflect the fact that the
oxidative pathway is the most effective way of transforming glucose into biomass.
The maximum glucose uptake rate is also the most influential parameter for the
prediction of the ethanol concentration (produced by reduction of glucose), during
this first growth phase. The glucose saturation rate plays an important role,
however not as significant as the maximum uptake rate (r G,max : SRC = 0.74;
K G : SRC =-0.48).
Obviously, the results of the global sensitivity analysis (SRC) should be
compared with the results of local sensitivity analysis (Fig. 2 ). It can be seen that
both methods rank the biomass yield on glucose (oxidation) as the most influential
parameter. For the ranking of the other parameters, there are quite some differ-
ences between the results obtained by the two methods.