Modelling Approaches for Bio-Manufacturing Operations - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 103

application of GSA has the potential to reduce the number of design variables, thus
limiting laboratory effort and cost.
5 Visualising Search Spaces
Once the critical parameters have been identified and evaluated through modelling
approaches, one needs comprehensible ways to understand the resulting data in
order to make decisions. Models can generate large amounts of information, and
although in certain cases simple graphs may suffice to understand the location of
an operating optimum, in other situations this may not be the case. The issue of
process interactions is a case in point; to develop operating policies that select
values of process variables that lead to optimal whole process performance, they
must account for the interplay between process steps and thus trade off the
potential beneficial effects of an upstream improvement against additional com-
plexities that this may induce further downstream. As indicated earlier, a classical
example of this considers a high-pressure homogeniser in which elevated pressures
or numbers of passes cause greater product release but at the expense of a more
heavily micronised debris that is more difficult to remove from the supernatant in a
following centrifugation step. Working out the best trade-off in conditions between
the two steps for yield and clarification requires use of suitable methods for
graphically identifying the shape and size of the search space. Hence there is a
need for approaches that can visualise outputs intuitively to identify combinations
of operating parameters that satisfy required performance levels. Methods such as
windows of operation can assist with this activity, in the homogeniser-centrifuge
case, for example, by selecting conditions which meet levels for both minimum
product recovery and debris removal. Thus, windows are a powerful way to extract
design information from quite complex mathematical models and so inform an
engineer about which conditions are process-relevant.
Windows of operation are formed by plotting critically important input
parameters on the axes and then applying chemical, physical, biological, process
or financial constraints to the search space to identify a bounded region that
simultaneously satisfies all threshold values for process outputs and product
specifications. Windows may be used to define feasible regions of either parts of
processes or sequences of steps. The method can be used during both initial
process design as well as for post-approval changes to identify operating bound-
aries and to judge process robustness in response to changes in constraints or
parameter ranges. Sensitivity analysis as discussed above can also be useful for
determining how the ranges of operating variables and their interactions may affect
the shape and size of the windows. Woodley and Titchener-Hooker [ 45 ] describe
how bioprocess design windows can be used to find acceptable operating regions
in both qualitative and quantitative forms. In the former case, theoretical infor-
mation can be used to get a rough idea of the characteristics of a search space,
while in the latter case and with the right models, input data and understanding of