Bioprocessing Europe Lisbon

Updated: Apr 10, 2020

Last week we have been to Bioprocessing Europe in Lisbon. The conference was organized in a brought manner, offering both, an up- and downstream processing stream, as well as a gene and cell therapy stream. With more than 100 experts presenting their current bleeding and cutting edge technology, the conference, hosted in the Sheraton Lisboa Hotel & Spa, was a great deal. We were also presenting the recent progress with our technology. While Gerald Striedner was introducing the latest updates in soft sensor development for upstream technologies, Moritz and I were highlighting the need to step away from static process descriptions towards dynamic process models.

Why should the whole dynamics of a bioprocess only be described at the endpoint of a process? Why not incorporate process variables over time and additionally use hybrid model approaches to also get a better process understanding from prior process knowledge. Let me bring a provocative comparison: For us, the current status in upstream development is somehow similar to cooking pasta! Why?

Why do we still describe dynamic bioprocesses with static approaches? The time to change has come!

If you look at the package of a certain type of noodles, it will tell you how long the pasta has to be cooked until the perfect quality (“al dente”) is delivered. However, the whole complexity of the process is neglected. It is just assumed that when you keep the water temperature constant, your pasta quality will hit the sweet spot after the described “process time”. But how is this different from bioprocess development? In upstream development, a lot of parameters are currently screened in complex design of experiments (DoE) worldwide. However, the product quality (titer, glycosylation patterns, host cell content, remaining DNA and others) is only described for the process endpoint. But what if process deviations occur? What if, to get back to the noodle example, the heating system fails? Your noodle quality will not be in the desired space, because the process itself is not understood! If you do process development with DoE you most probably have to repeat the experiment because you assumed that the temperature was kept constant.

Comparison of static process description with classic DoE and dynamic process description using hybrid models. While the classic DoE approach only evaluates a certain process at the endpoint, the hybrid model approach incorporates every time point. Hence, also process deviations are used to build the final model

So in our approach, we incorporate the process data and build a hybrid model based on the whole process. Hence, if a certain parameter changes over time (let us assume the temperature control fails for two to three hours) this information is used to build the model, thus also enabling model predictive control.

Apart from the four talks we presented, we also got the chance to share our insights in a workshop. Although we are constantly presenting our approach to dynamically describe bioprocesses on quite a lot of conferences, the feedback is still refreshing and we obtain a lot of requests of companies to help them set up those strategies.

Apart from the highly productive conference, also Lisbon is always worth a visit. With its hilly terrain, the historic center, the cable cars, and the Mediterranean cuisine it feels like a mixture of Rome, Vienna, and San Francisco. Whenever the next opportunity will arise, we will come back.

“Até breve Lisboa”

It's quite though to describe the beauty of a city in a single picture :) But I tried my best. Here you can see: Rua Augusta Arch which was built to remember the city's reconstruction after the earthquake of 1755