Data are vital for any industry 4.0 process, but the information needs to be of the right quality, says Will Hart co-founder of digital software consulting firm Opvia.
“There is great scope for significant advances in biomanufacturing with tools like machine learning. For this to happen, companies must ensure they are collecting all the data and metadata required to put their results in context,” he points out.
“Biology is inherently more complex than traditional manufacturing, so you must record more data and particularly metadata. Further still, the data are often of more complex types themselves making existing tooling insufficient.”
The challenge for most biopharma companies is how best to set up the technologies needed to collate information from multiple unit operations. And many struggle because they take an ad hoc approach, according to Hart.
“Building infrastructure to handle such quantities and complexities of data from scratch is incredibly difficult. Many companies try to reinvent the wheel and end up with something looking a bit square. It simply doesn’t make sense for everyone to be building their own systems from scratch,” Hart tells GEN.
Scope
Another difficulty is the range of information generated during biomanufacturing processes. Ensuring that critical quality attributes (CQA) tracked in a bioreactor work in context with parameter data generated during downstream processing can be a significant challenge for the inexperienced.
The best approach is to use an IT infrastructure capable of taking data from multiple sources on the line, enriching it. and building a useful digital model, continues Hart.
“There is a wide spectrum [of] data that biopharma equipment generates, ranging from simple formats and comprehensive to deliberately obfuscated and lacking in context,” he explains. “It’s absolutely vital to have a system that integrates, standardizes, and enriches the data generated by such equipment.”
Solutions
Despite the technical challenges, Hart is confident industry will find a way of standardizing the management of biomanufacturing data and be able to take full advantage of innovations like artificial intelligence (AI) and machine learning (ML).
“With the right foundations, AI and related methods could, and should, be staples of any all aspects of biopharma product development from discovery to manufacturing,” he says. “This change is already being driven by the increasing complexity and volume of data required to understand and innovate.”
“The longer this trend continues, the greater the need for computation tools to assist the scientist’s skill and intuition.”
For products for rare diseases or personalized medicines, machine learning, which is a form of AI able to automatically learn and improve from experience without being explicitly programmed, is likely to be of considerable benefit, adds Hart.
“ML is extremely useful for smaller runs,” he explains. “Typically, models can be trained on a wider set of data across different runs and start to generalize. This means they can provide a lot of insight into even the first attempt at a new process.”
