Biopharmaceutical Manufacturing is reorganized with digital twins and AI

Biopharmaceutical Manufacturing is reorganized with digital twins and AI ...

The quality of a drug product is determined by the manufacturing line. Preclinical assessments and clinical trials provide an indication of safety and efficacy, but it is the manufacturing process that ultimately determines how a drug will perform as a product.

Simply put, a poorly-controlled manufacturing process will result in product variability and put patients at risk. In contrast, a well-controlled manufacturing process produces a consistent product that can literally save lives. This uniformity is an essential component of the chemistry, manufacturing, and control strategies that manufacturers must demonstrate to regulators.

Clinical results are inadequate in terms of obtaining drug approvals, according to Mohamed Noor, the PhD candidate and digitalization manager at the National Institute of Bioprocessing Research and Training (NIBRT) in Dublin. Regulators will want to be assured that clinical manufacturing by a sponsor can lead to robust commercial manufacturing, batch after batch throughout the lifecycle.

The construction of a process control facility may be a difficult task.

According to Anurag Rathore, PhD, the director of the DBT Center of Excellence for Biopharmaceutical Technology in Delhi. It includes systems architecture, software applications, hardware, and interfaces, all of which are optimized and compiled per demand. This requires to be accomplished while keeping process requirements, production costs, regulatory limitations, and data acquisition in mind.

Process understanding is the foundation of any control strategy, according to ICH's Q8 recommendations.1 modeling is the best way to increase process understanding and meet regulatory quality-by-design expectations. The models should explain the relationship between process parameters and drug quality and performance attributes.

The most effective technique so far is to develop data-based models, scale-up, and process control. However, their predictive power is limited to the amount of data available, and they require considerable research work.

Traditional scale-down approaches to predict scale-up performance provide no system knowledge or physical understanding of the process, and are therefore susceptible to error when applied to make predictions during scaling-up.

Mechanistic modelassumptions based on established principles rather than data are becoming popular. Li explains that these models can provide a comprehensive description of the system, increase prediction power, and potential to extrapolate well outside of calibration space. They are useful tools to predict scale-up process performance, thus de-risking large-scale manufacturing operations.

Mechanistic modeling can help engineers better characterize and thus understandproduction processes, beyond scale-up. For example, they can be a valuable tool for characterizing and understanding design space.

Mechanistic models, according to Li, are high-fidelity models that allow users to simulate hundreds and thousands of experiments without extensive experimentation, revealing a far better picture of the design space.

She compares them to traditional central composition design techniques, such as central composition design, which selects only a few experimental points. Often, complex biological processes cannot be fully described through simple quadratic equations [even though such equations are employed] in central composition design.

Mechanistic models can be applied to process automation control systems, according to Li. They are useful throughout the product development lifecycle, including process optimization. She says they can improve process understanding, support quality-by-design process characterization studies, and risk large-scale clinical and commercial manufacturing runs.

Li and her colleagues on the Sanofis vaccine chemistry, manufacturing, and control team used a mechanistic approach to simulate a chromatography run. In addition, the model was accurately described as a digital twinhas. Besides supporting process scale-up, the model provided a proof-of-concept demonstration of how mechanistic chromatography might facilitate in silico process development and characterization.

Digital twins are a powerful part of process management. Parviz Ayazi-Shamlou, PhD, is the vice president of the Jefferson Institute for Bioprocessing in Philadelphia, where the twins are designed during development. These twins are crucial to process control and more.

Ayazi-Shamlou believes that creating a digital twin for a bioreactor would be beneficial for bioprocess engineers to develop and operate a cell culture operation entirely in silico. Once established, he explains, this kind of digital twin can be used to assist process development and optimization, answer what if? questions about the operation, and interrogate process deviations, significantly reducing experimental work, time, and costs.

He says that digital twins technology is used in training next-generation bioprocess scientists and engineers, almost like aviation simulators are used today in training pilots.

In some industries, the use of artificial intelligence (AI) in manufacturing has become the norm. Among the industries covered by a recent MIT survey2, AI is a third most common usage.

According to Kiefer Eaton, AI has found applications in biopharmaceutical production in a more limited manner. Generally, its applications are focused on improved process control of individual unit operations. These include bioreactor-centered or chromatography-column-centered operations that are responsible for the extensive downstream and downstream operations for a biologic manufacturing line.

In a recent research done by scientists, a manufacturer might recreate an entire bioprocess facility by combining the data from each unit operation in series and funneling model outputs between units operations.

Continually, a holistic approach might be used to identify bottlenecks in a production process or for more complete whole-plant optimization. For example, sacrifice local losses in product yield during an upstream process might lead to higher final product yield if downstream purification steps were improved (for example, by preventing aggregates or undesirable post-translational product patterns in the case of biologics).

The possibility of total-process AI control is significant. A key advantage over conventional data-driven approaches is the capability to more accurately model and understand a processeven in real time.

We at Basetwo, we strive to combine machine learning or AI with engineering knowledge to develop hybrid process models that can learn process dynamics better than traditional mechanistic or data-driven approaches. These models use the power of AI in an engineering context to extract process data.

These models will, according to Eaton, help manufacturers predict when the best to harvest or transfect a batch. It will also provide an alert when a batch or process is falling out of spec, giving engineers ample time to ensure corrective measures are taken before a deviation occurs or product yield or quality is compromised.

Cameron Bardliving, PhD, is a slightly different perspective. Although Bardliving believes that AI may play an important role in process control, he points out that there are still product-related challenges to overcome.

Bardliving describes the complex connection between biopharmaceuticals and their process environments as one of the greatest barriers to greater use of AI. It is challenging to predict priori a biologics critical quality attributes, and the lack of first-principle knowledge about the structure-function-process triangle makes it difficult to develop a fully predictive AI-based model that can be used in a manufacturing environment.

Although the introduction of full AI into biopharmaceutical manufacturing has been relatively slow, substantial progress has been made. Moreover, work continues to be completed across several levels to make a fully integrated smart AI-based manufacturing factory of the future a reality. Ultimately, the objective of any biomanufacturing process is to meet regulatory standards.

References to the European Medicines Agency, the European Medicines Agency, and the International Medicines Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use. ICH Guidelines for Pharmaceutical Development, June 22, 2017. 3.McCauley D. The global AI agenda: Promise, reality, and a future of data sharing. Trends Biotechnol. 2020; 38: 11411153. DOI: 10.1016/j.tibtech.2020.05.008.

You may also like: