Using Big Data Analytics to Optimize Commercial Chromatography
Traditionally, commercial-scale chromatography processes are monitored and reviewed using qualitative visual comparison of profiles against a reference batch. Big data analytics (specifically Multivariate Data Analysis, or MVDA) offers biopharma companies the opportunity to automate and digitalize chromatography review steps to optimize performance and commercial operations.
This article is posted on our Science Snippets Blog
Process verification and batch review pose challenges for many biopharma companies because they are lengthy and resource intensive. Using paper records without a data management system makes this process even more challenging.
During chromatography, a lot can go wrong, and there are huge stakes involved. Contamination, protein failure, or problems with the column packing could result in a lost batch or no recovery of the product, which is very costly – both financially and in terms of time. Moving toward a big data analytics approach to chromatography evaluation reduces the risks of batches gone wrong, helps identify the root cause of any problem sooner and allows for optimization of a process based on real-time monitoring.
Why Big Data Analytics for Chromatography Operations?
Why would a company want to move toward digitalization of its chromatography monitoring and use advanced data analytics? For several reasons
First, to allow for better process optimization and repeatability. Data analytics helps operators to detect (or even prevent) process deviations before traditional approaches might show a noticeable deviation. Optimization means being able to understand the reasons (or root cause) for cycle-to-cycle and batch-to-batch variations in order to improve cycle times, yield, and product quality. Repeatability means being able to implement golden batches (the ideal batch) in real-time by detecting drifts in membrane pressure, pH elution profiles, or other process variables, and receiving a warning before a drift starts to affect the process.
Second, to move toward digital transformation. Data analysis, including MVDA and DOE, have been used in upstream processes successfully for many years now, but in some companies, the downstream processes are still using paper and visual comparisons. In order to embrace digital transformation and biopharma 4.0, companies must move toward using data analytics in all processes.
Third, to enable continuous bioprocessing. MVDA enables continuous bioprocessing by converting large datasets into visual information. The FDA encourages manufacturers to use multivariate data models to improve processes, reduce variability and improve the consistency and quality of products. From quality by design (QbD) to continuous process manufacturing, regulatory agencies in the pharma and biopharma spaces have embraced and promoted the use of data analytics.
A data-based, digital approach moves your company toward a more quantitative approach to monitoring and builds a platform for repeatability by instilling an exact, measurable process. Process validation is an inherent part of quality control, and data analytics provides a robust methodology that quantifies your process.
What is Multivariate Data Analysis?
Multivariate Data Analysis, or MVDA, is a statistical technique used to analyze data that is generated from more than one variable. By creating summary variables, or a big picture that takes into account all the variables at once, MVDA helps you see the relationship between the variables so you can better understand cause and effect.
On the other hand, univariate analysis looks at just one variable at a time to evaluate the process status. This means you can miss the effect that variables interacting with each other may be having on the process.
If you’re just comparing batches one at a time or one factor at a time in the batch, you're not able to catch slow drifts or the interactions that are happening that might be creating deviations. Using MVDA with real-time process monitoring means being able to receive warnings before drifts start to affect the process and taking action to avoid costly mistakes. In addition, you can forecast how batches and cycles will finish based on the current trajectory of the real-time data.
MVDA can help companies address these challenges by enhancing their chromatography data review process. By using MVDA, small deviations and peak shapes can be detected, which is not possible with a traditional manual review. These are some of the reasons that a multivariate review is better than a univariate approach.
How Does MVDA Work?
Multivariate models from historical or live process data are used to monitor and analyze chromatography processes. The analysis takes into account all relevant process parameters. By using these techniques, the current process can be compared to historical runs that were successful and met quality requirements.
Common variables for chromatography models include:
- UV post membrane
- Temperature pre column
- Pressure pre membrane
- Pressure line before filters
- Flowrate after pump line
- Delta pressure membrane
Using MVDA, models can be built in a number of different ways, with phases or without, to monitor batches, batch evolution, or to compare batches. Examples of phases to use for chromatography data include equilibration, loading, washing, elution, and re-equilibration. Choosing the best method depends on the level of detail the end users expect from the models and how they intend to use them.
SIMCA® MVDA facilitates the simultaneous analysis of multiple chromatography phases and parameters by reviewing digital chromatograms. The tools makes it easy for non-data scientists to set up models and understand what’s happening with easy-to-read summary charts and drill-down charts.
After models are built using historical data, or reference data, a real-time process monitoring solution such as SIMCA®-online can be used to connect to the chromatography data from your system as the processes are running. Then you can quickly turn your chromatography data into information that summarizes all parameters, cycles, and batches.
With SIMCA® MVDA, multiple chromatographic phases and parameters can be analyzed simultaneously.
The Steps for Using Data Analytics
- Model the historical data. Use SIMCA® MVDA to create reference models based on historical process data (such as control data, analytical data, raw material data), which will help you expedite root-cause analysis and batch review investigations.
- Connect real-time data. Connect real-time process monitoring data to SIMCA® models through a data management system such as PI.
- Monitor in real-time. Monitor incoming batches with SIMCA®-online from a control room or the plant floor to identify any process deviations in real time.
- Share data online. Provide partners with real-time access to SIMCA®-online dashboards through the web client.
What Can MVDA detect?
Some Case Examples
Let’s take a look at how MVDA has been applied in the real world to support and optimize chromatography review.
Case 1: Evaluate Continuous Protein A Chromatography Performance
Traditional chromatography process monitoring, such as visually reviewing elution peaks, is time-consuming, error-prone, and not fit for continuous bioprocessing. Pall and Sanofi Aventis used SIMCA® MVDA to enhance chromatography monitoring and review, ensure performance consistency, and expedite process release.
MVDA can detect small deviations in peak shapes earlier than traditional methods can. Using SIMCA®, the company was able to:
- Monitor process consistency (cycle-to-cycle reproducibility)
- Detect column-to-column variations
- Detect column failures and other trends before they become problematic
Case 2: Expedite and Digitalize Chromatography Review
FUJIFILM Diosynth Biotechnologies in BioProcess International1 used SIMCA® and SIMCA®-online (with the PI System as the data backbone) to expedite their chromatography review process. By detecting deviations in batch consistency, cycle consistency, and other trends, SIMCA® can alert the user to impurities before they become a serious issue or, in the worst case, cause a batch loss. The benefits they achieved by using data analytics for chromatography review included:
- Review times were shortened
- Resource expenditures were optimized
- Paper footprint reduced by 10,000 sheets per year
- Their partners can access and share data on-demand
- Collaboration opportunities and increased trust with partner
Case 3: Ensure Smooth Transfer of Chromatography Operations to a CMO
When transferring processes to a contract manufacturer (CMO), it can be difficult for the product owner to know if their process is performing optimally. Using SIMCA®-online, Takeda was able to demonstrate that recovery and purification processes were comparable in real-time.
- MVDA charts help to visualize distinct groups of populations and can illuminate major differences in the process itself or the asset framework (i.e. equipment or instrumentation)
- Determine if batch-to-batch or site-to-site differences occur and identify the underlying reasons
- When transferring from one site to another, MVDA can be used as a tool for defining measures of equivalence
MVDA Enhances Chromatography Processes
The traditional method of monitoring chromatography processes, such as visually examining peaks, is time-consuming, error-prone, and unsuitable for continuous bioprocessing. By using big data analytics for chromatography evaluation, the risk of batches going wrong can be significantly reduced, the root cause of issues can be identified earlier, and process optimization can be achieved using real-time monitoring.
+ Download a Free Trial of SIMCA®
+ Get a Free Demo of SIMCA®-online
References
1. Jensen and Caroço. “Enabling Digital Chromatogram Review for a Faster and More Reliable Operation”. BioProcess International, 2020. https://bioprocessintl.com/sponsored-content/enabling-digital-chromatogram-review-for-a-faster-and-more-reliable-operation/
