(141e) Development of an End-to-End Data Management and Visualization System for Cell Culture Process Development | AIChE

(141e) Development of an End-to-End Data Management and Visualization System for Cell Culture Process Development


Doyle, B. - Presenter, Gilead Sciences, Inc.
Opel, C. F., Gilead Sciences, Inc.
Rusev, D., Gilead Sciences, Inc.
Lamadrid, I., Gilead Sciences, Inc.
Yeung, W., Gilead Sciences, Inc.
Derfus, G. E., Gilead Sciences, Inc.
Bai, Y., Gilead Sciences, Inc.
Cell culture processes generate extensive data sets originating from diverse systems. These data are collected on varying timescales, ranging from single metadata values to trends from on-line sensors generating thousands of data points per day. Additionally, data are stored in various formats and locations, making uniform retrieval and visualization challenging. Historically, we have used Microsoft Excel to manage data, perform simple analyses, and generate fixed visualizations. While Excel offers a familiar environment for handling data, it is limited in its processing capabilities, ability to directly compare differently sized or shaped data sets, and ease in flexibly creating and formatting graphics. Using Excel as a repository and analytical tool for comparison of data between experiments, bioreactor scales, laboratories, or manufacturing facilities requires an inefficient and labor intensive workflow of manual transfers and manipulations that is error prone, difficult to control, and time consuming.

To better maintain and use cell culture data, we developed an end‑to‑end data management and visualization system to meet user requirements while presenting familiar and user friendly interfaces that are accessible to both highly trained laboratory personnel as well as site leadership.

In the cell culture process development laboratory, the core of the system is a customized control system implementation for process control, coordination of run metadata, and automation of process execution. Users interact with simple templated Excel worksheets to deliver run parameters to the control system. A PI server archives data from the process control system and in-process analyzers; parallel data streams from pilot plant and clinical manufacturing campaigns are shared into the historian across enterprise level PI connections. PI data from active and historic runs across all scales can be queried and retrieved by PI client applications for real‑time visualization and comparison or ad‑hoc analysis. Archived data from all sources are processed through a routine to automate data interpolation, alignment, and calculations. Processed data is available in a relational SQL database for use by powerful visualization, business analytics, and analytical software.

This system has replaced Excel as the primary access point to cell culture data, reducing labor and risk of errors by automating the data management process. It has enabled flexible comparison and more valuable visualization and analysis of data across runs and scales which previously were not feasible. We envision this system to have applicability from process development studies to ongoing monitoring of manufacturing campaign data and to be expandable to be inclusive of process data from contract manufacturing facilities and product quality analytical data.