An Advanced Monitoring Platform for Rational Design of Recombinant Processes - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 69

bioprocess monitoring is heavily impaired by the lack of direct readable online and
in situ sensor equipment. However, there has been significant progress in online
signal acquisition and sensor technology as well as expansive development in the
field of offline biochemical and molecular-biology-based analytical equipment.
Therefore, the objective and the opportunity are to close the gap between the broad
spectrum of bioanalytical offline methods and online devices. This can be done by
the application of mathematical models, multivariate data analysis (MVDA), and
computer science to set up ''soft sensors'' [ 26 ]. These sensors can be based on
calculations utilizing online-accessible process variables or on correlations
between online and offline datasets. Such soft sensors trained on historical datasets
can be used to predict complex variables from previously unseen datasets by
statistical modeling, such as artificial neural networks (ANNs) or partial least
squares (PLS) (cf. chapters by Glassey and by Gernaey). In this context it is
important to note that the prediction is only valid for datasets within the previously
trained solution space; data cannot be extrapolated.
Moreover, the availability of these complex variables enables the implemen-
tation of control strategies that were previously not realizable. In addition to the
improved observability of the process, the availability of these complex variables
provides clues for the design of new control strategies.

1.2 Recombinant Protein Expression in E. coli Systems

For recombinant organisms the complexity of monitoring is even more pro-
nounced because there are additional variables that have a strong impact on pro-
cess performance. Genetic engineering techniques offer possibilities to modulate
growth and product formation independently from each other, and also provide
increased and subtle opportunities to develop process monitoring and control.
From a pure control perspective, the genetic features of recombinant host-vector
systems should allow improved controllability compared with nonrecombinant
production strains, such as inducible promoter systems allowing for control of the
recombinant protein synthesis rate. However, current development of recombinant
protein

production processes is widely empirically driven with restricted

predictability.
In E. coli the expression of recombinant proteins is generally achieved by placing
the foreign gene into a multicopy plasmid vector under control of a constitutive or
inducible promoter. In the early use of industrial recombinant processes, only weak
expression systems were available and even high-copy-number plasmids did not
enable full exploitation of the metabolic potential of the host cell.
Subsequently, genetic-based optimization strategies led to the development of
very strong and inducible expression systems such as the T7 system [ 27 ]. However,
these strong vector systems could not be used efficiently because the product for-
mation period is limited by metabolic breakdown of cells triggered by too high
recombinant gene expression rates [ 28 - 31 ]. Hence, the potential of the cell factories