Modelling Approaches for Bio-Manufacturing Operations - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 101

or operating costs, and modelling can help to determine the effects of these risks.
Stochastic modelling involves assigning probabilities to ranges for specified inputs
to enable a user to determine the range of possible outcomes and hence the
confidence one can have in a given outcome or the risk of failing to meet specific
process performance criteria. Modelling the random nature of a bioprocess or
market-related forces can help to predict the most likely outcome, thus forming the
basis of risk assessments and mitigation strategies (e.g. to ensure that market
demand is satisfied or to avoid going over budget). Following on from the Lim
et al. paper discussed in the preceding section, another study was used to assess
perfusion pooling strategies for a mammalian cell culture expressing mAbs. Monte
Carlo analysis was used to simulate random run-to-run changes [ 26 ] by averaging
the outputs achieved over repeated model runs. Hence frequency distributions
were created for the amount of mAb produced annually and the cost of goods. Risk
factors that were modelled included the chances of contamination caused by
having larger numbers of interventions over extended culture durations as well as
the impact of uncertainties in mAb titre and the cost of the broth media. As part of
this, the study also tried to identify a manufacturing strategy which used resources
most efficiently at commercial scale. Other variations which could be modelled
include resource costs, dosage level and market demand. A similar Monte Carlo
business-process modelling approach was used by Farid et al. [ 11 ] in a case study
designed to determine whether a start-up company should invest in either stainless
facilities or disposable equipment for making material for early clinical trials. The
impact of variable product demand and product concentration in the culture step
were analysed in order to quantify the attractiveness of different manufacturing
options.
Aside from Monte Carlo techniques, another way of determining the impact of
variability is to use sensitivity analysis techniques which determine the effect of
changing chosen input parameters upon the required outputs. This is discussed in
more detail below.
4 Sensitivity Analysis
The task of choosing which parameters to test in a model is complicated by the
plethora of choices. Flow rates, feed concentrations, residence times etc. are just a
few of the parameters that can influence the performance of many unit operations.
Although Ishikawa diagrams and FMEA-style analysis coupled to heuristics and
prior experience can help to narrow down the parameters to the most fruitful ones,
a more ostensibly quantitative assessment can also be invaluable. Quantitative
understanding about which input variables are the most significant can help to
focus modelling and thus ultimately experimental development efforts upon the
key attributes that control manufacturing performance. This can be particularly
relevant when considering interactions between inputs that can complicate the
selection of bioprocess operating conditions. In a simple model with a clear