Genetic Engineering & Biotechnology News

NOV1 2018

Genetic Engineering & Biotechnology News (GEN) is the world's most widely read biotech publication. It provides the R&D community with critical information on the tools, technologies, and trends that drive the biotech industry.

Issue link:

Contents of this Issue


Page 55 of 57

18 | SEPTEMBER 1, 2018 | enabling technology that is penetrating every aspect of modern life. The science that drives precision medicine would not be possible without advanced data analytics. The application of big data analytics to biomanufacturing is still in the developmental stages. Beyond real-time control and re- porting, automated solutions enhance the ability to collect and analyze histor- ical product, process, operational, and performance data. Analysis of this data can result in several useful outcomes, such as predicting equipment failures before they happen to allow sched- uled maintenance, as well as control- ling process values to optimize product yields. The Big Leap: Machine Learning The big leap forward will come with the utilization of big data to de- velop and validate analytic models that can predict and optimize process pa- rameters and sequencing, resulting in better product yields. GE Healthcare has developed scale-agnostic machine learning (ML) infrastructure to address yield-efficiencies in bioprocess produc- tion facilities. With select biopharma clients, the GE Healthcare data science team has used advanced ML algorithms to produce process-parameter values designed to predictively maximize yield of each production run, developing so- phisticated ML/AI algorithms to drive efficiency and improve product yields in biologics manufacturing. For this purpose, GE Healthcare data science teams, in collaboration with on-site manufacturing teams, col- lect all relevant bioprocess data: these datasets include online bioprocess (per minute) data, manufacturing execution systems (MES) data, laboratory infor- mation management systems (LIMS) data, raw material manufacturing and QC data, manual entries, and informa- tion about corrective and preventive actions (CAPA) for all bioproduction scales used to manufacture biologic therapeutics. A proprietary data-inges- tion pipeline then converts, digitizes, and standardizes the datasets. ML models are then trained on all noncorre- lated input features with labeled output parameters such as batch yields and/or viable-cell-density values at each step of the biomanufacturing process. Directed by the ML models, fea- ture-set correlation analytics typically results in the reduction of the input parameter sets from several hundred to a few dozen, specifically down to the few dialable parameters. This produces actionable input values that the opera- tor can dial into the process to achieve high-yield runs on a regular basis. Even with the expected production variance of raw materials, GE Healthcare's ML models typically demonstrate accuracy metrics of greater than 90%. This end- to-end rigorous approach produces ML models that can reliably and ac- curately predict batch-specific yield ef- ficiencies. Data analytics in bioprocessing will continue to grow in importance. Much of the innovation associated with auto- mation technologies relates to the cap- ture, communication, analysis, and utilization of data. Bioprocessing will incorporate technologies drug makers are confident will improve their busi- ness objectives, including those with either no regulatory impact or those that provide improved regulatory com- pliance. Stable, robust technologies that reduce or eliminate risk without intro- ducing risk of their own will be critical. Continuous improvement driven by Big Data will drive greater efficiency and higher yields. The future of process op- timization and cost reduction involves new players—your automation team, your data science team, and your exist- ing process historian database. References 1. CPhI Pharma Insights. Prospects, Analysis and Trends in Global Pharma. CPhI Annual Report 2017. report 2. GE Healthcare Life Sciences. An environmental life cycle assessment comparison of single-use and conventional bioprocessing technology. c4ae463b8c4bd7594802fa5d/19083-source/op- tions/download 3. Radspinner D et al. Can A Shared Biomanu- facturing Facility Be the Answer to The Demand Forecasting Dilemma? Life Science Leader. August 13, 2018. Single-Use Technologies for Bioprocessing Evolution of Single-Use Bioprocessing into BioPharm 4.0 Continued from page 17

Articles in this issue

Links on this page

Archives of this issue

view archives of Genetic Engineering & Biotechnology News - NOV1 2018