Multiscale Experimental Design for Automated Bioprocess Development

• Applicant: Dr.-Ing. Mariano Nicolas Cruz Bournazou
• Subject Area: Biological Process Engineering
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 561489561

The main objective of this project is to develop a framework to tackle the Optimal Experimental Design (OED) problem over several stages of bioprocess development: a Multiscale Experimental Design (MED). High Throughput (HT) laboratory experiments are a key technology to accelerate Research and Development (R&D) in biotechnology, biopharma, and bioengineering. To date, research institutions and companies around the world use these technologies to speed up drug discovery as well as strain screening and engineering, process development, and more. Similarly, analytic robotic facilities have developed fast, achieving medical and biotechnological analyses by thousands in parallel. The use of robotic laboratory systems, miniaturized cultivation systems, microfluidics, design of experiments, and the integration of in silico models has brought upstream and downstream bioprocess development significantly forward in the last decade paving the way for a new era in bioprocess development. We can roughly group the main stages of bioprocess development into: 1. selection of host, 2. strain screening, 3. selection of operating conditions, 4. process design, and 5. scale-up. Each of these stages has unique difficulties, ranging from process engineering to microbiology. In fact, the experiments are often performed in very different laboratories, even located far apart from each other. For this reason: 1. it is difficult to analyze the process holistically; 2. it is also very hard to develop bio-processes effectively; 3. automation of the complete developmental pipeline is very challenging; and 4. the creation of ad- equate computational tools is hampered. A thorough overview in literature related to bioengineering and bioprocess development shows that there are no mathematical methods nor algorithms reported that support the “optimal” transition procedure between different stages of the developmental pipeline. Thus, there exists no method to compute a multiscale experimental design campaign. It is even difficult to find publications on bioprocess development that consider multiple experimental stages at once. Existing methods for Optimal Experimental Design for example, only offer methods to plan and perform experiments in each one of the stages (e.g., screening, benchtop, etc.). As a result there is no methodology to support very important decisions in the developmental process, as are: when to move from one stage to the next one; or what is the expected gain of information of each stage at the current developmental state. Formulating this problem is not straightforward as it involves several considerations that are difficult to embed in standard optimisation algorithms. Additionally, the quantification of important decision support variables (e.g., Expected Information Gain (EIG)) are neither simple nor cheap to compute. Finally, even experts make these decisions based mostly on pre-defined protocols or rules of thumb.

• DFG Programme: Research Grants