- How should we vary the controls (e.g. the time profiles of
feed flow rates)?
- When should we take the measurement samples?
Based on an optimality criterion (for example D-optimality)
experiments are specifically designed for the determination of the
significant parameters (vector
2
k p
) and once the experimental data is
available the values for parameters
2
k p
are determined explicitly,
yielding vector
exp
k p
.
Step 4. Range of validity [52]: The predictive capability of the
refined model needs to be tested against a set of independent
experiments with varying environmental conditions. If the model is
not able to describe the experimental trends satisfactorily then a new
model of increased fidelity needs be developed and the whole process
needs to be repeated.
Step 5. Model based optimisation [57]: Dynamic optimization
techniques can be used to identify worst- and best-case scenarios for
the operation of fed-batch and continuous cell cultures. The former
should be avoided and the latter needs to be further refined
experimentally so as to reach truly optimal operating conditions. Of
particular interest are dynamic optimisation techniques that can
remain computationally inexpensive whilst dealing with model
uncertainty and scarce sampling times.
The use of model-based techniques can facilitate the reduction of
unnecessary experimentation by indicating the most informative
experiments and providing strategies to optimise and automate the
process at hand. The presented research approach attempts to
integrate modelling, experimental design and validation with model
based control and optimisation within a closed loop framework, that
leads to increased productivity and reduced production costs for cell
culture systems. The integration of these four research tools represents
an elegant interdisciplinary approach that addresses the complicated
research and industrial problem of model-based control and
optimisation of cell culture processes.
Traditional models of microbial growth kinetics are based on the
assumption that description of the rate-limiting step produces an
adequate description of the process. Therefore, the Monod model,
which is perhaps the best classical description of growth kinetics, is
based on the assumption that culture growth is limited by a single
rate-limiting enzyme reaction following the well-known Michaelis-
Menten kinetics [23]. However, although traditional models can be in
some cases very accurate, they are apparently not capable of capturing
the regulatory effects controlling upstream the production of catabolic
enzymes, providing a rather simplified and idealised view of complex
biological processes [58]. The current progress in molecular biology
can be used to unravel the underlying biological mechanisms that
regulate gene expression and cellular function. High-throughput
experimental technologies are able to elucidate the behaviour of a
biological system at a holistic level. The results generated are known
as ‘omics’ data and constitute of genomics, trascriptomics, proteomics
and metabolomics, which measure gene, transcript, protein and
metabolite profiles of cells [59]. During the past few years the
advances in the ‘omics’ technologies have facilitated better
understanding of the function of microorganisms as industrial “cellfactories”.
This recent ability to acquire mechanistic knowledge of cell
function at local and global level enables the replacement of empirical
models with mechanistic ones, thus advancing the development of
efficient bioprocess models for industrial biotechnology [60].