The 2019 Umetrics User Meeting drew more than 102 engineers, operations managers, process experts, researchers, and data scientists in industries ranging from biopharma to food and beverage to chemicals who gathered to share ideas and insights into new methods for streamlining their processes, reducing waste and cost of goods sold.
The Umetrics 2019 User Meeting focused on helping users find ways to get more value from their data with highlights including the launch of SIMCA® 16, new innovations in product integrations and artificial intelligence (AI) applications.
Users and experts in data analytics presented case studies and new approaches for using multivariate data analytics, real-time monitoring and continuous process improvement to achieve better results. They explained how they have discovered innovative and more efficient processes, and uncovered surprising secrets in their company’s operations by using advanced data analytics.
Some of the topics presented by users and experts included:
- Real-time multivariate statistical process monitoring in biopharmaceutical manufacturing
- Predictive analytics models for large molecule manufacturing
- New monitoring models that combine continuous and discrete variables
- Leveraging real-time process information and advanced analytics to support QA/QC
- Using data analytics to predict patients at risk for adverse treatment reactions
- Applying deep learning and artificial intelligence (AI) to advanced data analytics
The annual user meeting helps manufacturers, researchers and data scientists find ways to get more value from their data by looking at the latest trends and methods in multivariate data analytics.
This year’s meeting was held at the Sartorius headquarters in Goettingen, Germany, May 8-9, and included a tour of the facilities. Attendees got to visit a production site and see operations supported by the Umetrics Suite of data analytics solutions — including real-time monitoring — in action.
Robert Soeldner from the Sartorius Corporate Research BioProcessing Lab presented the unique features of the BIOSTAT® Rocking Motion fermentation bioreactor, a wave-mixed bioreactor used for regenerative medicine, mAbs, vaccine development, and seed train. During the facilities tour, meeting attendees got to see it first-hand, and experience how the Umetrics® Suite of data analytics solutions supports more efficient operations and decision-making.
The user meeting tour included a live demonstration of the Sartorius Fermentation Bioreactor BIOSTAT® RM amplified by the power of Umetrics® Suite data analytics solutions.
The BIOSTAT® RM bioreactor is used for a high cell density perfusion cultivation supporting cell densities of more than 100 million cells/ml. The RM Perfusion bag features a built-in perfusion membrane, which makes the use of an external cell retention device obsolete. Perfusion is controlled gravimetrically by integrated loadcells for level control as well as an external harvest balance to control the perfusion rate.
In addition, Sartorius CEO Joachim Kreuzburg presented the Sartorius plan for new development and stressed that in the light of Industry 4.0 adoption, Sartorius remains highly committed to the growth of data analytics tools across industries, including chemical, and food and beverage manufacturing.
Meeting attendees also had the chance to speak with Umetrics Experts, to ask question, see demos and get answers to specific challenges they have in data analytics.
Updates from Umetrics® Suite
Meeting attendees heard from Umetrics® Suite product managers about the latest product updates and the planned improvements in future releases.
A highlight was the launch of SIMCA® 16, which was released May 8, with product updates focused on an improved user experience including new ribbons, tours, wizards, data merging, multi-block analysis and more.
The much-anticipated Umetrics® Suite Roadmap was also rolled out at the user meeting. Key feature upgrades planned for the future include:
- SIMCA®: web clients and out of context data (2019); spectroscopy, data infrastructure and directed workflows (2020)
- MODDE®: Design Advisor, design space wizard, optimization wider (2020)
- Active Dashboard: enhanced CPV, more KPIs, continuous processes (2021)
Umetrics also introduced some innovative new products that resulted from collaborations with other companies and departments, including Sartorius Stedim Biotech. These new products are designed to bring robust data analytics tools into the hands of business users or researchers, not just data scientists.
Ing-marie Lindström, Umetrics product manager, explained that as user groups for data analytics expand toward “citizen data scientists,” a more persona-based approach is needed, along with guidance, automation and application specific workflows. With this comes more need for empirical data based on Deep Learning, which can turn unstructured data into structured data.
One example is a new software tool for ambr® users focusing on data analytics for the clone selection process. The ambr® clone selection tool simplifies the clone selection workflow and helps to make the decision-making process more consistent. Previously, the clone selection process could take 20+ weeks per project with researchers using univariate analysis rather than applying advanced statistical software. The new application puts reliable clone selection analysis in the hands of non-data scientists.
Another software innovation coming soon is a tool that supports scale conversion throughout the drug development workflow. The Sartorius Scale Conversion Tool will support translation of process parameters between bioreactors, adjustment of recipes between scales, enable bioprocess knowledge to be used to inform scaling, and enable scaling to inform risk assessment. It is scheduled for internal release in June 2019 and commercial launch in 2020.
The Sartorius Scale Conversion Tool is scheduled for commercial launch in 2020.
Further exploration of how artificial intelligence and deep learning can be applied to amplify the results of data analytics came in the final keynote. Prof. Dr. Andreas Dengel, DFKI (German Research Center for Artificial Intelligence) presented “Artificial Intelligence as a Means for Intelligence Amplification and Augmentation.”
Two other highlights of the meeting included keynote speakers from Amgen and Baxter.
CENK UNDEY, Executive Director of Process Development at Amgen presented “10 Years into Real-time Multivariate Statistical Process Monitoring in Biopharmaceutical Manufacturing using SIMCA-online: Reflections and Future Prospects.” He discussed how his company has continued to develop their use of data analytics to increase efficiencies and agility of processes and product development across all plants.
“Data is the new oil. You need to refine it to increase its value.”
- CENK UNDEY, Amgen
He explained what success looks like at Amgen:
• Saved batches. Contamination avoidance in bioreactors. Early leak detections and bioreactor vent heater failure early detection.
• Yield increases. Titer increase due to induction timing, seeding and batch production optimization.
• Efficiency gains. Faster non-conforming resolution due to ease of access to data and ability to compare historical batches. Downtime avoidance, quick trouble-shooting, preventive/condition-based maintenance.
Lee Hutson, Quality Director for Baxter, a global healthcare company that focuses on renal care, advanced surgery, pharmaceuticals, infusion systems, nutritional formulations, and acute therapies presented “Multivariate Data Analysis (MVDA) at Baxter.” Baxter began working with advanced data analytics in 2004 and now has more than 30 manufacturing sites across the globe that use MVDA. He charted their journey from using MVDA in a reactive state to using real-time data to be more proactive to implementing predictive models.
Since globalization of their MVDA program, they have completed 200+ installations, implemented 100+ process models, generated 60+ MVDA success stories, and realized significant cost avoidance savings across all operations.
Process Industries Presentations
Focusing on other industries, two well-received presentations were from BASF in the chemical industry and Syngenta, in the agrochemical industry.
David Hajnal of BASF presented: “Combining Statistics, Engineering Art, and Science: Tailored D-Optimal-Designs for Chemistry Lab.” He said that BASF develops, produces and sells a large variety of materials that need to be customized for changing customer needs. Many lab experiments are needed during product development and optimization of chemical recipes and processes. This leads to challenges in designing and processing experimental data.
Using MODDE Pro DOE software for Constrained D-Optimal Design helps with maximizing variance, determining technical feasibility and using prior knowledge to inform new experiments. He said that Constrained D-Optimal Design is an important game-changing tool for synthesis and formulation of chemicals, such as adhesives, wall paints, industrial coatings, corrosion protection, performance materials, etc.
Mark Earll in the GBJH Product Characterisation Group at Syngenta presented: “Applications of Chemometrics at Syngenta.” He gave a very comprehensive presentation on using multivariate modeling of tomato metabolite data and for developing mosquito insecticides.
He presented four different varieties of tomatoes to compare three non-ripening varieties with a normal (wild) tomato during fruit development. The team compared “time based OPLS” to “design based OPLS” where different experimental design factors were encoded. By comparing the OPLS loadings for each genotype vs. time, they were able to observe metabolite changes that are unique and are shared by each genotype.
The team concluded that chemometrics is a fantastic tool for typical wide and noisy research data, and that click and point software such as SIMCA is essential for data mining. Mark stressed that the benefit of visualizing data cannot be underestimated. He is a firm believer that multivariate methods provide a toolkit for wide research data, especially to help with:
- Interpretable models
- Computationally efficient
- Handling missing data intelligently
- Estimating predictability
- Modeling pessimistic models for yields that are unlikely to over-fit
- Enabling pattern recognition, data mining, regression, classification
Making data work harder
From hearing first-hand accounts of others user’s success with MVDA, to new ways to use data to predict outcomes, to optimizing experimental designs using DOE, to finding ways to apply examples from other industries to their own, attendees at the Umetrics User meeting found interesting and trending topics and plenty of chances to learn from peers.
Want to know more?
Download the user meeting presentations now.