A new Quality by Design (QbD) approach in early bioprocess development

One of the main challenges in pharmaceutical manufacturing is how to quickly characterize a new strain or clone for robust bioprocessing along the product life cycle. Process development groups in industry as well as in academia share a common task: Strains and clones are quickly generated by molecular biology. However, it remains a major question which one to choose for developing a robust process and how to avoid failures during scale up and along the product life cycle. A further challenge is that proper strain/clone characterization shall be based on strain/clone-specific physiological parameters (Key Performance Indicators, KPIs) that describe the strain/clone independent of scale and initial conditions. 

An ideal system would be an integrated experimental setup allowing for easy experimentation, representative sampling and proper analytics as well as objective data analysis to derive strain/clone specific physiological parameters. To develop such a system, experts from academics (Zurich University of Applied Sciences/ZHAW) and industry (PreSens, Exputec – now part of Körber’s Business Area Pharma) partnered in a unique collaboration with a promising result.

Status quo with several downsides

Most experiments today use shake flasks which are the most common type of cultivation systems in early bioprocess development. This makes possible conducting a large number of convenient and affordable experiments in parallel. Typical usage contexts for this setting are early process design steps like screening a process design for strain characterization, process optimization and medium development as used in academia, large-scale industry and start-ups.

While offering the advantages mentioned above, this setting also brings with it some major challenges: Sampling here tends to be error-prone as oxygen-saturation drops due to rotation stop, volume decreases and lost data points. The calculation of characteristic values for strain characterization is human-biased. Due to different calculation methods and documentation, values might not be comparable, also because offline and online measurement data sometimes are sourced from different systems. As a result, data is documented in varying Excel files and annotations are not consistent. Also, the amount of generated data can be overwhelming, and processing and selecting this data volume is time-consuming. As a further downside, shake flask experiments are often restricted by oxygen limitation, which leads to non-comparable values.

Automated process design and intuitive data evaluation

In the context of the joint development project Intellishaker, funded by Eurostars, the partners, PreSens, ZHAW and Körber developed a completely new experimental environment. The goal was to avoid the disadvantages of traditional shake flasks settings and achieve at least the same affordability and convenience.

Our unique solution provides real-time, continuous and non-invasive measurements using optical sensors (PreSens) and allows for the prediction of critical events, like oxygen limitation or the best time for harvest with a sequential Monte Carlo Method. This machine learning approach, more precisely, a particle filter prediction with adaptive co-variances for each prediction step, keeps the forecast flexible in case of changes in metabolism (figure 2). Another benefit is the automated and clearly defined calculation of physiological KPIs for strain characterization. A KPI table (figure 3) facilitates intuitive data evaluation.

Figure 3: An example of a KPI-table: For each experiment (first column from the left) the main values for characterization are calculated. qO2 is the specific oxygen uptake rate, mumax shows the maximum specific growth rate for different growth models (Chmiel, Gompertz, logistic) and from direct parameter-extraction calculation from the particle filter (paramex). The two columns with y display the yields.

Our new QbD approach results in several clear benefits for pharma manufacturing processes:

• Quick and objective strain characterization 
• This avoids failures in scale-up and along the product life cycle and speeds up process development and time-to-market.
• The setting is applicable for every organism that is following a typical growth.
• It provides an intuitive software environment that is user-friendly and guarantees simple and result-oriented everyday usage.

This next level of automated process design and intuitive data evaluation is an excellent example of teamwork, uniting experts from academics (ZHAW) and industry (PreSens, Körber), and is the tangible result of a holistic approach for shake flask experiments.

Learn more: Quality by Design web seminar

To gain more insights, join our Quality by Design (QbD) web seminar on Thursday, 24 September, 16:00 CEST.

Barbara Pretzner from Körber’s Business Area Pharma together with speakers from ZHAW and PreSens will talk about “Accelerating clone and strain selection using shake flasks, online monitoring and advanced data science - novel QbD approaches in early bioprocess development.”

Register now!