Editorial: Bioprocess Development in the era of digitalization

Peter Neubauer Florian Glauche M. Nicolas Cruz-Bournazou In the last few decades, a vast number of research findings in biotechnological processes were reported with high commercial relevance. Still only a small fraction of these molecules was successfully brought to industrial production. The development of production processes in biotechnology still requires a lot of time and money compared to other industries. Due to the complexity of biological systems, the commercialization of cell-derived products needs a combination of deep scientific knowledge and a thorough understanding of process engineering. In addition, regulatory requirements must be met, which may increase the overall risk of success. Process development in biotechnology is still mainly driven by experienced personnel creating and evaluating experimental data. However, in recent times, automation, miniaturization and data science are setting new paths for drastic changes in bioprocess development. Today, hundreds of automated experiments can be performed in parallel miniaturized cultivation and purification systems per day. Additionally, statistical experimental planning and evaluation is applied to utilize the experimental capacity of these facilities efficiently. Still, to exploit the full potential of automated laboratories, innovative software concepts and workflows are needed. Moreover, the conditions in which small-scale experiments are performed need to resemble the production scale as closely as possible 1. Today, there are numerous possibilities to create, collect, store, and share data. This has a significant potential in biotechnology where: i) the complexity of living organisms can be studied deeper and analyzed by computer aided tools, ii) complex biochemical pathways including hundreds of thousands of reactions can be reconstructed, and iii) proteins can be designed using in silico tools and gene synthesis. Nevertheless, while today´s development of processes and products in other areas of engineering can be performed strongly on the basis of computer based models with a minimum of practical experiments, the future bioprocess development is still coupled to extensive wet-lab experimentation due to the nature and the complexity of bioprocesses. The most significant challenge is that it is difficult to create a reliable mathematical model of even the simplest prokaryotic cell. Changes in protein expression, metabolic activity, mutations, etc. can neither be foreseen, described, nor deeply understood. In other words, model based methods in biotechnology will always be coupled with experimental activities to validate and re-adjust models to the reality of living systems. From advanced plant-wide control and optimization strategies for bioprocesses to custom made therapeutics, fully automated experimental facilities are required to run thousands of experiments in parallel to produce large amounts of data needed to constantly validate the mathematical models. These robotic laboratories must be intelligent when designing experiments and learning from the data. Today's laboratory automation systems aim rather at an increased experimental throughput than at performing complex intelligent operations. However, for making bioprocess development more efficient, these systems must be equipped with adaptive experimental design methods, to be able to plan and perform complex experiments and e.g. decide when to take representative and informative samples and to initiate the analysis. Failures should be detected automatically and promptly, the experimental strategies should be re-adjusted as data is being generated, and the experiment should deliver the maximal amount of information possible. In the near future, the user should be able to enter some specifications in the computer (e.g. operating conditions, clinical status, available substrates, desired product) which will be used to: i) find candidate organisms, ii) design screening experiments, iii) carry out the experiments, iv) analyse, evaluate, and store the data, and v) send the specification of the new process (including strain, required facilities, operating conditions, and control strategy) for its production. Methods for Design of Experiments (DoE) have been developed and widely applied to this end. Unfortunately, in biotechnology we are often dealing with systems in which time dependent responses are important (e.g. industrial processes, patient response to medication). Therefore, the experimental design should consider the nonlinear dynamics of the process, in which case more advanced adaptive tools such as Model Based Design of Experiments (MBDoE) are additionally required. These methods have to be further developed, aiming to i) design optimal dynamic experiments in very short time, ii) communicate with the robotic facility that preforms the experiment, iii) learn from the data as it is being generated, iv) re-fit the parameters of the model to the data, and v) re-design the experimental strategy, while keeping the biology of the system in focus. The building blocks for such a facility already exist and are being developed further: computer capacity, interconnectivity, modern non-invasive sensor technology, robotic facilities and operating routines, efficient dynamic databases, methods for experimental re-design, and advanced optimization programs. These blocks need to be integrated to create an intelligent laboratory for rapid characterization of biosystems including: easy interaction of all devices, experimental design programs tailored for robotic LHS, online optimization algorithms that enable an optimal operation of the robots, and advanced control methods to assure the correct operation of the facilities. This special issue "Effective Bioprocess Development" covers recent advancements in the field and showcases several innovative approaches to open questions. The selected papers discuss and indicate the feasibility of how far bioprocess development can be automated by the integration of modern liquid handling stations, cultivation devices and analytical instruments. The article by Herold et al. reports the successful application of a progressive parameter-control workflow during process development of a lactase production process. With this strategy, the time frame for process development was reduced to 3 months 2. Tailored software solutions for the integration of model-based techniques are presented in the article by Wellenbeck et al. 3. One of the key elements of the efficient use of automation in biotechnology is the organised storage of all data in a database, as described in this issue by Schmidt et al. 4 in a way that these data are immediately accessible for e.g. evaluation and re-planning of the experiments. An example of how "intelligence" can be incorporated into such a complex system where the whole experiment is not only performed, but also digitally controlled is described by Nickel et al. 5 (see also Cruz-Bournazou et al. 6). At a somewhat lower level automatic operations also can be used to identify important parameters for process development, such as the dependence of the specific product formation rate on the specific growth rate – an important information for establishing a fed-batch process 7. When the amount of data is limited, hybrid models can be applied as an alternative to pure statistical analysis for better process understanding, as demonstrated by von Stosch et al. 8. Moreover, a summary of the different tools for process development and a perspective on the efforts in linking up- and downstream process development are given in a review by Baumann and Hubbuch 9. The issue closes with the report of a DoE based strategy for efficient automated cell disruption 10. The broad range of topics illustrates the interdisciplinary nature of bioprocess development and highlights the necessity of combining these innovations into future process development platforms. If molecular and cell biologists, software programmers, process engineers and business administrators work closely together, production efficiency, time-to-market and costs of bioprocesses and new bio-products can be reduced significantly.