Remember meForgot password?
    Log in with Twitter

article imageQ&A: Data science and biotech team up for success Special

By Tim Sandle     Jun 20, 2018 in Business
Few companies these days are aiming to reduce costs of pharmaceuticals. To meet this ambitious goal, biotechnology company Biogen teamed up with data scientists OSIsoft, in relation to an Alzheimer’s drug.
Biogen needed to rethink their biologics manufacturing, with more effective data operations monitoring and quality control. To achieve this, they teamed up with OSIsoft. OSIsoft’s expertise enabled Biogen to manage and process its manufacturing data with full visibility and context all the way to the production floor.
The two companies also worked together to iron out some process variabilities. Part of the variability was in the overall yield of the drug, which was typically at 5 to 7 percent. With current tools, a company might only be able to test a drug after it’s produced. However, the use of analytics and real-time monitoring (what’s termed Process Analytical Technology) allow for in-process testing and verification.
This approach provided Biogen with the ability to take early samples and take corrective measures so they do not need to throw away the batch. In relation to this paradigm, and as an example of how companies can work together with data sharing, Biogen were able to leverage some of the testing from the supplier so it can accelerate its own testing without compromising quality, and while reducing the process to one day.
To discuss how these process improvements were achieved, and how data science can assist biotechnology in general, Digital Journal spoke with Petter Moree, Pharmaceutical Principal for OSIsoft and Tim Alosi, Head of Global Data Analytics for Biogen.
Digital Journal: What is the current state-of-play for Alzheimer’s treatments?
Tim Alosi, Biogen: Alzheimer’s is the sixth leading cause of death in the United States and more than 25 million suffer from the disease.
PET scans showing the differences between a normal older adult s brain and the brain of an older adu...
PET scans showing the differences between a normal older adult's brain and the brain of an older adult afflicted with Alzheimer's disease.
National Institutes of Health
DJ: What is Biogen’s goal in relation to Alzheimer’s?
Alosi: Biogen is a life sciences company that focuses on neurodegenerative diseases. Our goal with our upcoming product is to serve 1,000,000 patients with Alzheimer's by 2020.
Petter Moree, OSIsoft: Biogen is rethinking all aspects of their biologics manufacturing for their new drug Aducanumab, from improved testing and analytical instruments to advanced controls and predictive models, which is where OSIsoft comes in. Their goal is to reduce the cost of the drug by 80%.
DJ: How did you team up with OSIsoft and what is the objective?
Alosi: Biogen has worked with OSIsoft for a number of years but recently has collaborated more closely with the company to start using the data collected by the PI System in a more fundamental way and give Biogen a ‘single version of the truth’ so to speak that can improve production.
Moree: By teaming up with OSIsoft and leveraging our PI system, they’re using data and predictive analysis to implement a new manufacturing process that can lower yield, decrease chances of variability, and at the same time, lower the cost. Biological processes and fermentation are more challenging to produce than conventional medicines. As natural processes, they can't fully be controlled, thus causing variations and deviations. Biogen is reducing the impact of variability by applying analytics early in the fermentation process to correct deviations. If the company can see a particular batch is going out of range, it can shift processes to bring it back in line. Currently, a company may not be aware of deviations until two days into the process.
The PI system captures operational data from machines or “things” and synthesizes it so engineers or data scientists can see solve problems before they arise or optimize production flows to increase productivity and reduce costs.
A substantial element of reducing the cost is increasing the yield of active materials per liter. Think of it as a way to improve capital utilization: The higher the yield, higher the productivity and output. 5 to 7 percent is the typical industry yield for biological pharmaceuticals. Biogen in testing of its new reactors is up to 12% right now, and are aiming toward 15 percent. In the very early days of the industry, biologics used to cost $10,000 per gram. Biogen will be at $100 to $150 a gram when they hit big volumes.
DJ: What does the OSIsoft PI system provide?
Alosi: If you want to do analytics and data you must have strong, accurate and consistent context. I’ve heard from others that 70 to 80 percent of time is usually spent on managing data before they even start their analytics. If you don’t have context, that’s the reality of it. If we can break this barrier, we can unlock a lot of value in our data. PI system does a great job with presenting asset context and creating one common context of asset and temporal context.
Our goal with OSIsoft’s PI is to turn it into the central point of information driving some of our really key data-rich applications and work processes. It largely goes back to that element of context and not just time-series data. It’s the two together that enable these applications.
We’ve chosen to implement an enterprise-wide PI system because of cybersecurity. We want to drive as much access to PI data as possible, but we have to balance that with our requirements around cybersecurity. We could potentially face an attack at any time so we need to protect our assets at the plant level.
So how do we balance that? We’re using our enterprise agreement to leverage an enterprise-level historian that we use to aggregate all of our tag data and our AF data with synchronization up and down. We’re currently beta testing the connector in order to aggregate event data also.
Data is the new oil.
Data is the new oil.
DJ: What does the PI system do in terms of data?
Moree: The PI System essentially captures and organizes the vast number of data streams being generated by manufacturing equipment, sensors and other devices so people can use it to save energy, spot problems or do things like making “smart” products. Overall, you could say it provides two functions: machine data storage and organization (similar to the storage management systems in IT) and a decision support console that helps employees conduct analytics.
Like other pharmaceutical companies, Biogen’s main concern is quality: it wants to ensure that output conforms as close as possible to its “golden batch” formula and reducing variability. Through real-time monitoring and analytics with the PI System, Biogen can see if there are issues early in the manufacturing process, and take corrective measures to fix them so they don’t have to throw away batch. These can help both quality and cost.
Manufacturing data can be used in other ways to cut costs. When it comes to warehousing, 4 to 7 weeks might elapse between when a truck drops off raw materials and it gets released for production. With data sharing, Biogen can leverage some of the testing from the supplier so it can accelerate its own testing without compromising quality – while reducing the process to one day. Fewer days mean less costs in warehousing and internal testing. At the other end, Biogen wants to reduce having to carry 40 days of inventory of finished products and move toward JIT.
Additionally, the PI system is helping reduce compliance processes and eliminating all paper. With PI, a lot of redundancies and administrative burdens can be reduced. Biogen expects it can reduce data integrity reviews and audit trail review by 70%. Audit trails and reviews still happen, but the cost of doing them will be much lower.
DJ: How does this help with lowering drug development costs?
Moree: From the most immediate and tangible, the PI System can help cut costs by streamlining tasks like regulatory compliance and shipping on one hand and improving the yields and output on the other. It helps with both opex and capex in other words.
Additionally, the PI System is also used to reduce the cost of developing this drug. The PI System provides real-time data to reduce the time it costs to start analyzing experimental results. It can also provide insight into what steps of the drug development can be accelerated.
DJ: What types of metrics are useful here?
Alosi: Our first stages of testing only included 165 patients – now we’re aiming to test the results to about 2700, to eventually lead us to our goal of 1 million.
With this new manufacturing process, we hope to accomplish a 50% reduction in waiting time for results; a 70% reduction in batch exceptions; a 78% reduction in batch review time; and a high performing, robust and adjustable process
Moree: It really varies by customer. Overall, quality and consistency rules. Is a manufacturer achieving high, consistent, desired yields? Second, cost.
DJ: How are you measuring success?
Alosi: Every day and every batch counts. We can’t avoid losing batches because of the amount of throughput going into the facility. We need to reduce the limitations in legacy manufacturing to ensure that time and money isn’t spent completing tasks that a data management system can do. Currently, we’re measuring things such as yield, lead time, review time and output as pillars for success.
Moree: It’s subjective by customer, but we often go by customer satisfaction and growth. Are they using the PI System in more creative and expansive ways than when they started? And we’re happy to say the answer is generally yes. We serve 23 of the top 25 manufacturers and many if not most have moved from starting with functions like data management to improving various facets of production.
More about data science, Alzheimer's disease, Biotechnology
More news from