(470a) Accelerating High-Throughput Experimentation with Templated Statistical Analysis

Authors: 
Albrecht, J., Bristol-Myers Squibb
Rosso, V. W., Bristol-Myers Squibb
Saurer, E. M., Bristol-Myers Squibb
Chiou, G., Bristol-Myers Squibb
Roberts, F., Bristol-Myers Squibb
Janey, J., Bristol-Myers Squibb
Tabora, J. E., Bristol-Myers Squibb
Mack, B. C., Bristol-Myers Squibb
Automated high throughput experimentation for process development allows researchers to rapidly evaluate the effects of process conditions on the performance of synthetic steps. At BMS, an established workflow is used to conduct highly automated experimental designs. The final step of the workflow involves merging the automated reaction conditions with sample chromatograms (HPLC, GC, or LC-MS reports) and analyzing the data using statistical software. Often this is a bespoke, manual process that may be challenging to reproduce when the need arises in the future.

This presentation reports the development of a statistical analysis template that automates much of the repetitive, manual work of data analysis. Through a web interface, statistical analysis reports are reproducibly created from the raw sample data and shared with the project team. The work of merging data files, empirical model building, plotting trends in the data, response surfaces, and suggested future experiments are performed automatically with a minimal amount of user input. In addition to standardized data products, project teams have the flexibility to customize the analysis to address specific hypotheses. This template has been successfully applied at BMS to both automated and manually executed experimental designs.