As scientists and engineers work toward improving bioprocessing, one part or another—something upstream, downstream, or connecting streams—often gets the attention. At the 15th Annual Bioprocessing Summit in Boston, though, Richard D. Braatz, PhD, the Edwin R. Gilliland Professor, Chemical Engineering at the Massachusetts Institute of Technology, described a “fully instrumented testbed … for the end-to-end integrated and continuous manufacturing of monoclonal antibodies.”
Optimizing such a system depends on analytics. Putting together the right analytics package, however, often hits some snags. As Braatz and his colleagues reported in November: “For many biopharmaceutical manufacturing processes, mechanistic models are not available due to the lack of complete biological process understanding and analytical quantification.” In those situations, bioprocessors must collect data and then develop an analytics model around that information. Yet, this is a complex task.
Alternatively, a bioprocessor can pick an off-the-shelf data analytics/machine learning (DA/ML) software package, but Braatz and his colleagues noted that the “optimal selection of DA/ML tools requires a substantial level of expertise due to the diverse nature of biomanufacturing data in terms of both quantity and quality.” Selecting the wrong tools can lead to inaccurate predictions, which can take a bioprocessor down a path of making bad decisions about controlling a line.
Smart process analytics
So, Braatz and his colleagues tested out smart process analytics (SPA) software on their set-up for manufacturing monoclonal antibodies. Braatz’s team noted that SPA software “automatically selects DA/ML tools for manufacturing data based on specific characteristics of the data and expert domain knowledge of the process.”
The best DA/ML tools will depend on the process being modeled. Nonetheless, Braatz and his colleagues concluded: “The capability of smart process data analytics software to capture product- and process-specific characteristics enables wide application of the software to biomanufacturing processes.”