(423f) Interactive Prior Knowledge Assessments to Predict Process Performance

Interactive Prior Knowledge Assessments to Predict Process Performance

Paul Gramlich, Kyle McElearney, and Nitya Jacob

As program timelines accelerate to meet patient needs, it has become
increasingly important to leverage historical data sets to facilitate process
development and characterization.  In Amgen
Drug Substance Process Development these historical data sets were captured in
review paper-style documents focusing on individual cell culture and
purification unit operations. 
Unfortunately, these documents were extremely cumbersome to update and
were quickly outdated.  Additionally, as
static documents, they could not be customized for critical upstream and
downstream process attributes.  Recently,
a comprehensive effort was undertaken to mine critical process data from
hundreds of documents in both Electronic Lab Notebook (ELN) and paper
formats.  Data was compiled in Amgen’s
custom ELN software and consolidated in the data lake, were Tibco Spotfire^©
dashboards were designed to enable process development scientists to fully customize
prior knowledge data to predict the performance of any unit operation.  This innovative new system has already
dramatically streamlined FMEA scoring during process design planning; insights
from historical data that used to involve review of dozens of documents can now
be gained in a matter of minutes.  Overall,
this approach is a powerful example of the impact data science is having on the
biotechnology community’s ability to rapidly advance molecules to support
patients in need.