Biomedical Engineering Reference
In-Depth Information
As the precision of PLS improves with the number of input signals, PLS cannot
benefit from selected input signals.
The lowest root-mean-square errors of prediction (RMSEPs) for the selected set
of target offline variables were obtained when applying a nonlinear RBF model
that used selected online signals of dielectric and optical spectroscopy as inputs
(Fig. 4 ). The availability of versatile datasets from the different analytical devices
enabled the individual assessment of each dataset with regard to the quality of
prediction of the particular variable. In addition, benchmarking of different sta-
tistical approaches could be carried out. To apply the previously created model for
online prediction during a cultivation process, a MATLAB TM function was pro-
grammed to perform data input, chemometric modeling, and display of estimated
results online during a cultivation process [ 38 ].
Recently, an additional analytical device, the proton transfer mass spectrometry
(PTR-MS) for analysis of VOCs, became available. This instrument fits perfectly
into the monitoring platform due to its continuous and noninvasive measurement,
which allows more accurate assignment of the acquired signals to individual
compounds than the signal derived from 2D multiwavelength fluorescence and
dielectric spectroscopy [ 16 ]. Datasets from such analyses will further contribute to
prediction quality.
4 Process and Systems Design
Optimization of recombinant protein production in E. coli can be obtained by
either engineering- or genetic-based solutions. In retrospect, it is obvious that
concepts for cultivation of cells (e.g., high-cell-density cultivation, fed-batch
cultivation with diverse feed profiles, or chemostat cultivation), including process
monitoring and control strategies, were successfully developed in the first period
of recombinant protein production. After this period, progress in process engi-
neering slowed down because of limited monitoring capabilities and fragmentary
understanding of the biological systems. In this situation, genetic-based solutions
became dominant drivers for higher yields in recombinant protein production.
Today, a multitude of very powerful expression systems are available, but in many
cases the capacity of the cell factory is not only exploited but overstrained in these
systems. As a consequence, cultivation of such overbred strains is far more
demanding, process stability and reproducibility are reduced, and in many cases,
product quality negatively affected. The logical conclusion is that further opti-
mization calls for harmonization of measures in process engineering and host cell
engineering.
In this section, such an integrated process optimization approach is presented.
The frequently used T7-based E. coli expression system serves as a model system.
A typical example of the output of a standard process is shown in Fig. 5 .
Induction of recombinant gene expression initiates a very high recombinant
gene expression level. In parallel, a drastic increase in plasmid copy number from
Search WWH ::




Custom Search