Stem cell therapy promotes the use of stem cells in tissue restoration. These stem cells are grown in a lab and manipulated to become particular cell types, including heart muscle cells, blood cells, or nerve cells. Stem cells can either repair or replace cells damaged by disease or chemotherapy, or they can work with the immune system to combat certain cancers and blood-related illnesses like leukemia.
The cell therapy manufacturing process has several unique challenges, including a supply chain that is highly personalized for individual patients.
The procedures in the manufacturing process are very clear and must be carefully followed. The patient initially requests compatible CBUs (cord blood units), and after a suitable batch has been located, it is sent to the manufacturing facility. The CBUs are immediately placed in liquid nitrogen, and all necessary materials for production must be present before the process can begin.
Since the process cannot be interrupted from day zero until the harvest, free production slots must be located in the most efficient manner. The product is frozen after the final day of production (harvest) and then prepared for transportation to the hospital and the patient.
During manufacturing, work can only be done on one batch at a time in the clean room, which leads to potential bottlenecks.
A producer of stem cells was preparing to launch their new manufacturing operations. So, they contracted two consultants, Logico and Opyflow, to help validate the manufacturing site capacity and maximize clean room utilization.
The two consultants worked together to develop a solution using a digital twin for the manufacturing process. Since demand in this industry is very stochastic, time dependent, and has many variables, simulation was chosen over a simple analytical model.
The developers decided to use AnyLogic to create the digital twin because AnyLogic provides both agent-based and discrete-event simulation modeling. In addition, AnyLogic provided the capability to define and modify extremely complicated processes using Java programming to create and customize the Libraries. Importantly for this project, optimization could also be performed. And finally, a UI that featured 3D visualizations and KPI dashboards could be developed.
The required parameters that were needed for the digital twin were available to the developers from the physical asset. These included engineering, operations, planning, and quality. Additional parameters can be found within each of these, as shown in the diagram below.
The customer received a stand-alone model with adjustable parameters. In the simulation model, the process starts when raw materials arrive at the manufacturing facility. All auditing and operational procedures connected with these raw materials also take place at this moment.
Also, within the model, the staff arrive and go from an unclean area to a sterile area. All work is then completed in what is known as the clean room.
In each clean room, work is done on only one batch at any given moment. So, there could be the capacity to have a larger workforce, but the room can only be used for one batch at a time. Therefore, there is a need to maximize the clean room's potential and remove bottlenecks.
A multi-step optimization process was implemented for the purpose of identifying bottlenecks. The consultants used this approach to identify a bottleneck, correct the parameter causing the bottleneck, and then "lock" it so that it couldn’t be changed. They then repeated this process, and step by step they reduced the bottlenecks. This process could continue until all potential bottlenecks have been corrected.
According to the findings, there was potential for a 30% increase in capacity, which would result in 30% more sales. The client ultimately makes more profit.
Additionally, the client was better prepared for the launch because numerous issues were found and planning problems that hadn't been thought of were identified and resolved. An example of this was when, after identifying bottlenecks, extra equipment was ordered to clear them.
Finally, the digital twin for manufacturing and the FMEA (failure mode and effects analysis) were used to produce the paperwork that had to be submitted to the FDA. This was required to obtain a biological license.
Looking further ahead in this cell therapy manufacturing process, the potential for model expansion is limitless, but the first step will be to create a model that incorporates the entire supply chain.
The case study was presented by Yossi Benagou of Logico, and Dov Amor of OPYflow, at the AnyLogic Conference 2022.
The slides are available as a PDF.