Figure 4: Performance of the buffer management system during 11
days of operation of a continuous antibody production.
The requirement on quality analysis can be handled almost in the same
way as the buffer management system. As discussed above the continuous
downstream process could assume the conditions in the harvest tank to be
almost constant or changing very slowly. A set of performance and
quality attributes were measured through chemical analysis of samples
taken from different positions in the process flow path once a day. The
sampling of the flow path were performed by the automation system but
the chemical analyses were performed off-line and the actual attribute
were calculated after the process were completed. The quality analyses
were not used for monitoring of the operation, which would had been
beneficial.
In lab-scale continuous downstream process there is no need to sample
the flow path more often than a couple of times per day for any upstream
based disturbance, based on the slow upstream dynamics compared to
downstream. Assume that an HPLC analysis of a quality attribute takes 30
min and it is performed twice a day, then the HPLC is idle 23 hours
every day. A general analysis system like HPLC can be configured to do a
set of different analysis. If an analysis takes, in average, 30 min to
perform it means that an on-line HPLC can handle almost 50 analysis per
day. In lab-scale development resources are limited and an on-line HPLC
needs to handle all on-line analysis in the processes. One such system
was implemented and studied in Tallvod et al 2023 and 2022B. It is
called quality analysis system, QAS. The physical configuration of the
QAS was composed of two setups, one for sample handling and preparation,
and a second dedicated analysis system, here a HPLC. The preparation
setup handled the sample from the process and performed needed
preparation, like dilution, conditioning buffer exchange and chemical
modifications. The second setup, the HPLC, were configured to perform a
set of different analysis. The automation part of the QAS is composed of
three parts; i) sampling of the processing flow path, ii) queue system
for orders of needed analyses iii) prepare the sample for analysis, and
iv) preform the HPLC analysis and send data to database, see Figure 5.
The client application procedure does the necessary operations to do the
actual sampling of the processing flow path. This sampling operation can
be performed on demand based on an external order or by the client
itself. The sampling can be a small amount directed to a sample loop or
a complete pool shortly stored in a container, which is then pumped into
a sample loop, before further processing in the process. The size of the
sample depends on the analysis and the number of differ of analysis
needed. Each sample point in the downstream process has a collection
loop on the preparation system. When the sampling of the process flow
path is completed, an order for analysis is sent to the QAS queue
controller.
The queue in the queue controller, in an orbitX controller, is a
first-in-first-out, but it is possible to rank the analysis in an
importance order. Particularly if a time-critical-analysis is need for
on-line control, which means that the analysis needs to be performed to
generate a result in a given timeframe. The queue controller adds new
orders into the queue and removes orders that are completed.
When the preparation/analysis unit is ready, it gets the first order in
the queue from the controller. When the unit has performed the analysis
it communicates that the analysis has been done and the queue controller
removes the order from the queue. Note that the order can contain
multiple analysis for different attributes. The generated data from
preparation and analysis is sent to the Orbit database for further data
processing. In some situations, it is convenient to do the data analysis
locally, which is straightforward to do as a computation extension in
the local Orbit controller.