The use of scale-down models as accurate, robust, reproducible process development tools can be very challenging. In a newly-published eBook, the bioinnovation team here at Cambridge Consultants reveals how we take a systematic, robust approach to solve accuracy and reproducibility issues within scale down-models. Meanwhile, I have taken the opportunity in this article to set out a concise overview of the content, including a summary of our case study.
Creating robust bioprocessing scale-down models
Scale-down models have become an increasingly common development tool, especially in the bioprocessing industry. A scale-down model is where you take a large-scale process (such as bulk fermentation) and build a physically much smaller version so you can more easily do experiments to understand and improve the process. Theoretically, they can significantly de-risk product development by reducing costs and increasing speed. They remove the need for early investment in specialist equipment which may subsequently prove irrelevant, for example, and can shorten development timelines. As a result, a good scale-down model can support innovation and reduce time to market, thereby improving competitiveness and market share.
But small-scale model design is notoriously difficult. While it’s accepted that a model’s results will always represent some level of compromise, it is a real challenge to produce results which can be reliably scaled back up – and if they cannot, this then leads to loss of development momentum with associated budget and commercial impacts.
Material composition in bioplastic production
With reliability of scaling so important – yet often so elusive – there’s clearly a need to determine why failure is common and how it can be systematically addressed. To explore this question further we developed our own exemplar small scale-model and investigated its replicability. Our focus was material composition in bioplastic production, and in particular the production of polyhydroxyalkanoate (PHA). This type of bioplastic is bio-derived and biodegradable and is used in a wide variety of industries including medical, automotive and consumer goods. Naturally produced by a variety of organisms, specific PHAs – with controllable composition and properties – can also be engineered using industrial microorganisms. Our small-scale model was designed to produce PHAs to a defined specification using fermentation processes.
While it was relatively straightforward to produce two common PHAs, tuning the composition was less straightforward. We soon established the break point at which our ability to predict composition failed, thereby preventing the reproducibility of our model at different scales. We hypothesised that this was due to the way that different fluid dynamics (at these different scales) affected oxygen transfer rates and culture mixing and hence growth characteristics and PHA composition. In response, we deployed a multidisciplinary team, of biologists and fluid dynamicists to investigate further. Through fluid mechanic modelling and experimental verification, we developed a mathematical tool designed to help us predict the reproducibility of each small-scale model iteration.
For a range of small geometry vessels we found the volume of fluid, and shaker speed, which would generate results which could reliably predict large scale performance, and while it was satisfying to determine a scientific solution, our study also underlined two important truths about small-scale models. Firstly, small-scale models are not simply smaller clones of their scaled-up counterparts, and any variation can dramatically affect scale up reliability. Secondly, a wide skill-set is required in order to pinpoint, understand and address unreliability when discovered and then to study and define the critical process parameters which need to be refined. This breadth of knowledge is not always available to highly focused product development teams but should be factored into the development of any small-scale model, and certainly invaluable when problems become apparent.
We are now using our predictive model as a tool to help our clients design better scale-down models from the start. Using a systematic and mathematical approach, such tools can identify those experimental parameters at greatest risk of change when scaling up, and which – following more thorough investigation – can be adapted or redesigned, maintaining the relevance of the scale-down model’s results. This, in turn can help support the design of more robust and reliable small-scale models, better able to meet the key objectives of increased development speed and reduced cost. Please email me if you’d like to continue the scale-down conversation. It would be great to hear from you.