Drug firms seeking to make manufacturing as efficient as possible must be willing to hand some parts of process development over to machines, according to a leading researcher.
Automation has been used in the drug discovery lab for decades in areas like molecule design, synthesis control, and screening. Similarly, in the production plant, automation is increasingly being used to control production.
In the process development lab, by contrast, automation is little used. Developing a production process involves multiple experiments designed to determine how each parameter or input can be optimized to ensure the resulting product is consistent and in line with quality goals.
At present, most process development teams conduct experiments manually. This approach is time-consuming and costly and would be more efficient if automated, says Robert Schmitt, PhD, professor at the Laboratory for Machine Tools and Production Engineering (WZL) of RWTH Aachen University in Germany.
“Automation in process development would help in developing and optimizing processes faster and more efficiently, as automation allows for parallelization and thus a higher throughput of experiments can be conducted.
He adds: “The largest benefit of automation will likely be its combination with digitalization of bioprocess development as well as manufacturing leading to improved management of data and easier and better insight into process data.”
Challenges
Drug companies that do opt to automate process development will need to foster close ties with technology suppliers according to Schmitt, who says mapping out the specific requirements of the process being developed is vital.
“The success of an automation project is very much dependent on setting a concrete goal and communicating well with technology providers in order to deploy a lab automation solution that actually fits the bioprocess challenges well.”
A solid data infrastructure is also critical, according to Schmitt, who says the volume of information generated during high throughput experiments will need to be managed and analyzed as efficiently as possible.
“Increased lab automation often leads to larger amounts of process data being generated and stored. If this data is well prepared and all relevant meta-data is stored with it, this can yield great insights into the bioprocess.
“AI is already widely used to process and evaluate microscopic images of lab processes at our institute [Journal of Biological Engineering, Processes, and Computers in Biology and Medicine], this information can be utilized to make process decisions, such as feeding or harvesting cells in culture,” Schmitt says.
And, like automation, AI has potential application in process development, Schmitt says, citing challenging areas like cell therapy manufacturing as an example.
“We also see applications of AI in broader aspects of cell therapy manufacturing, such as resource and production scheduling. It may also help in simulating bioprocess parameters, i.e., cell growth to predict ideal harvesting time points and make educated decisions during manufacturing.
“The largest pitfall here is the lack of regulatory standards and guidelines on the development and deployment of AI, which is greatly needed for broad acceptance and trustworthiness of AI applications in medical use cases.”
