strategies. In the field of biotechnological production of basic chemicals and
biocatalysts, the reactor volumes are getting steadily larger to ensure the necessary
yields. Newly conceived processes must be robustly feasible on a large scale [1]. On
the other side, in the field of biopharmaceutical production, product yields in the
reactor originally had a lower priority for originator products compared to the costs
incurred in downstream processing. The expiry of many patents and the development of generic or biosimilar products has led to a price pressure that companies can
only counteract with very efficient and less variable bioprocesses. Additionally, the
implementation of single-use strategies, especially in connection with the lower
power input in single-use bioreactors, introduces scaling phenomena, i.e. imperfect
mixing issues already at much smaller scales, i.e. in reactors with 1–5 m
3 [2]. Still,
the main reason for the increased interest in scale-down investigations are failures,
lower yields, higher batch-to-batch variability, or even changes in the product
quality of processes performed in industrial scale. The difficulties to predict the
outcomes at this scale, the large number of unexpected responses of cells to different
environments, and the impossibility of computer based tools to foresee the changes
on the phenotype throughout scales based on laboratory data are the driving forces
behind scale-down popularity [1].
The discrepancy between laboratory- and industrial-scale fermentation processes
results from heterogeneous environments in large-scale bioreactors due to the
limited volumetric power input and geometric issues, which result in longer mixing
times associated with the increasing volumes at industrial scale, compared to
laboratory scale bioreactors [3]. The effects of such process inhomogeneity on
microbial physiology and product syntheses, including the quantity and quality of
recombinant products, have attracted much attention in the bioprocess research
community, due to the mostly unforeseeable impacts of the heterogeneities on
process efficiency, e.g. by the accidental incorporation of non-canonical amino
acids into the product [4].
In the past three decades, various forms of single-compartment and multicompartment scale-down bioreactors have been developed to study scale-up effects
in fermentation processes [3]. This development, however, is accompanied by a
constant discussion about the extent, to which these systems really reflect the
conditions at an industrial scale. Concrete proposals for procedures for scaling
down a process to laboratory scale have only been developed in recent years, see,
e.g. [5], but are generally very sophisticated and therefore unsuitable for broad
application.
In parallel, during the last years, there has been a phenomenal increase in the use
of high-throughput (HT) miniaturized bioreactor systems for strain screening and
bioprocess development, which has significantly reduced the times required for early
bioprocess development. These new powerful laboratory tools require, however,
new methods for planning, performing, and evaluating these highly parallel experiments. The systems are no longer treatable by manual methods – therefore, standard
methods of design, mathematics and statistics, modeling and process engineering as
they have been used in other disciplines for a long time have to be implemented and
adapted in the field of bioprocessing. Intuitively, when dealing with large data sets,
Potential of Integrating Model-Based Design of Experiments Approaches and. . .
3
biocatalysts, the reactor volumes are getting steadily larger to ensure the necessary
yields. Newly conceived processes must be robustly feasible on a large scale [1]. On
the other side, in the field of biopharmaceutical production, product yields in the
reactor originally had a lower priority for originator products compared to the costs
incurred in downstream processing. The expiry of many patents and the development of generic or biosimilar products has led to a price pressure that companies can
only counteract with very efficient and less variable bioprocesses. Additionally, the
implementation of single-use strategies, especially in connection with the lower
power input in single-use bioreactors, introduces scaling phenomena, i.e. imperfect
mixing issues already at much smaller scales, i.e. in reactors with 1–5 m
3 [2]. Still,
the main reason for the increased interest in scale-down investigations are failures,
lower yields, higher batch-to-batch variability, or even changes in the product
quality of processes performed in industrial scale. The difficulties to predict the
outcomes at this scale, the large number of unexpected responses of cells to different
environments, and the impossibility of computer based tools to foresee the changes
on the phenotype throughout scales based on laboratory data are the driving forces
behind scale-down popularity [1].
The discrepancy between laboratory- and industrial-scale fermentation processes
results from heterogeneous environments in large-scale bioreactors due to the
limited volumetric power input and geometric issues, which result in longer mixing
times associated with the increasing volumes at industrial scale, compared to
laboratory scale bioreactors [3]. The effects of such process inhomogeneity on
microbial physiology and product syntheses, including the quantity and quality of
recombinant products, have attracted much attention in the bioprocess research
community, due to the mostly unforeseeable impacts of the heterogeneities on
process efficiency, e.g. by the accidental incorporation of non-canonical amino
acids into the product [4].
In the past three decades, various forms of single-compartment and multicompartment scale-down bioreactors have been developed to study scale-up effects
in fermentation processes [3]. This development, however, is accompanied by a
constant discussion about the extent, to which these systems really reflect the
conditions at an industrial scale. Concrete proposals for procedures for scaling
down a process to laboratory scale have only been developed in recent years, see,
e.g. [5], but are generally very sophisticated and therefore unsuitable for broad
application.
In parallel, during the last years, there has been a phenomenal increase in the use
of high-throughput (HT) miniaturized bioreactor systems for strain screening and
bioprocess development, which has significantly reduced the times required for early
bioprocess development. These new powerful laboratory tools require, however,
new methods for planning, performing, and evaluating these highly parallel experiments. The systems are no longer treatable by manual methods – therefore, standard
methods of design, mathematics and statistics, modeling and process engineering as
they have been used in other disciplines for a long time have to be implemented and
adapted in the field of bioprocessing. Intuitively, when dealing with large data sets,
Potential of Integrating Model-Based Design of Experiments Approaches and. . .
3