(759a) A Case Study of Data Analytics for the Manufacturing of a Monoclonal Antibody

The U.S. biotechnology sector has
had double-digit growth rates in recent years (Imarc Group, 2012). In 2012,
sales of biologics were approximately $63.6 billion, with monoclonal antibodies
(mAbs) representing the largest fraction of this market with approximately 39%
of sales (Aggarwal, 2014).  Mathematical modeling of the manufacturing process
is one possible way to both support the growing biologics market as well as
decrease costs via improved control and understanding of process operations.
Modeling can play an important role in understanding, controlling, and
optimizing the process steps used in these processes (Tziampazis &
Sambanis, 1994).

In this case study, analysis
techniques are demonstrated in both bench-scale as well as process-scale
datasets from the production of a monoclonal antibody product. Bench-scale data
are analyzed using response surface methodology as well as regularization
techniques. Process-scale data are analyzed using correlation analysis, latent
variable methods, and regularization techniques. In all cases, this work shows
that careful attention must be paid to model calibration and validation,
particularly in biopharmaceutical applications where the amount of data is
often small. This work also demonstrates that, where available, unit operations
should not be isolated, but instead processed in sequence. Most importantly,
this work demonstrates that there can be predictive value in modeling critical
quality attributes using only a small, heterogeneous dataset. References

Aggarwal,
S., 2014. What's fueling the biotech engine - 2012 to 2013. Nature
Biotechnology, 32(1), pp. 32-39.

Imarc
Group, 2012. Global Biopharmaceutical Market Report and Forecast
(2012-2017), s.l.: International Market Analysis Research & Consulting
Group.

Tziampazis,
E. & Sambanis, A., 1994. Modeling of cell culture processes. Cytotechnology,
Volume 14, pp. 191-204.