(483c) Advancing Quality by Design Through Diverse Uses of Latent Variable Methods In Vaccines and Biologics Manufacturing

It is generally understood that latent variable multivariate modeling methods have great value-add potential in vaccines and biologics enterprises. What is relatively less-known, however, is the diversity of tasks that these tools can support in such applications. Although the fundamental math and statistics behind these tools is fixed, each of these different tasks require unique considerations with respect to several attributes, including data alignment/integration, data preprocessing and scaling, model validation, computing hardware, and visualization of results.

This paper will use recent case studies to illustrate the various ways that latent variable methods support current vaccines and biologics efforts at Merck, which include QbD initiatives.  These case studies will include retrospective analyses of existing operations for troubleshooting, exploratory modeling of processes for continuous improvement efforts, multi-scale models for scale-down model qualification, and consistency monitoring of manufacturing operations.