How Embedded Data Analytics Improves Raman Instrument Success
Supporting Instrument Manufacturers with Embedded Data Analytics Solutions (SIMCA®-Q)
Embedding data analytics solutions into hardware or connecting directly with other applications allows manufacturers to develop instruments that meet the growing need for robust process development, supporting PAT applications like Raman spectroscopy.
This article is posted on our Science Snippets Blog.
Spectral data is increasingly becoming part of the discovery, development, and quality management toolkit in pharmaceutical and biotech development. Being able to incorporate robust, validated data analytics models right into lab instruments gives manufacturers an advantage in creating products that meet the needs of today’s digital data-driven marketplace. For example, data analytics solutions can be embedded in or connected seamlessly with instruments for spectroscopy, NIR, Raman, mass spectrometry or chromatography, or integrated directly into a bioreactor.
For pharma companies, using a quality by design (QbD) approach favored by regulators to create consistent, validated processes requires reliance on strong statistical models, data analytics, and dependable data management. Embedding data analytics solutions into hardware or connecting with other applications offers many advantages for scientific discovery and process development that relies on spectral data such as Raman.
Rather than trying to build data analytics models and tools yourself (which is likely outside a hardware or instrument manufacturer’s area of expertise), you can incorporate an already developed, tested, and proven multivariate engine that integrates seamlessly with your preferred software. As an embedded data analytics solution, SIMCA®-Q simplifies and speeds instrument development, while offering users access to the gold standard of data analytics: multivariate data analysis (MVDA) using SIMCA®.
Why Use Multivariate Data Analysis with Raman?
In biopharma development, Raman spectroscopy can replace hours of difficult or time-consuming manual measurements and provide very useful information about a process (or quality of an ingredient or final product). But analyzing and making sense of all the data, and understanding how the samples being studied have affected or were affected by a process, is very complex.
Multivariate data analysis (MVDA) provides a proven, regulatory-approved method to uncover patterns, trends, and outliers, and detect cause and effect relationships. SIMCA® is an MVDA software solution that helps scientists organize data, extract the required information, understand the relationships between different data points, and visualize it.
Data analytics can help scientists understand “why” something is happening in a process. With SIMCA®, users can drill down on an outlier and see why it’s happening.
MVDA is a useful (even essential) tool for evaluating Raman data. The complex nature of Raman and other types of spectroscopy data means that there isn’t just one measurement point, but instead, an entire spectrum of data. Handling such complex types of data – and being able to parse out the important signals from the “noise” in order to understand the how or why – requires very complex, interconnected, and well-documented algorithms and models.
SIMCA® (and SIMCA®-Q) can help make sense of the data, separate the relevant information from the irrelevant, and create a clear picture of causes and effects. This involves building a model, testing it, and using it to measure (or even predict) future data progression.
One advantage of using SIMCA®-Q is that Sartorius has packaged everything that Raman suppliers need to incorporate MVDA into your products: algorithms, pre-processing ability, and predictive capabilities. The SIMCA®-Q Raman tools are part of a predefined package that is very easy to embed and connect with your hardware or instrument.
Creating Robust Raman Models
When you build a multivariate model, it can be used for reporting, predicting, and forecasting a process. That means historical data provides input to create a model that can predict – in real-time – how a process is playing out (providing advance warning that will allow users to take action early if a process isn’t performing optimally).
Historical data can come from different setups, such as bioreactor status or spectroscopy data. In the example below, the data comes from a Sartorius bioreactor (either small scale or up to 2000 liters or more) equipped with instrumentation that provides data such as oxygen mixture, temperature, PH, or other processes measurements.
Spectroscopy Data Feeds Model Creation
You can also get additional information from a Raman probe or another type of spectroscopy. Models can be built as an overview or to create more specific setups such as using soft sensor spectral information about concentrations or titer. It can be for diagnostic or prediction uses (and in automated systems, for control).
Data generated in real-time (such as from Raman probes) can be used to predict how the process compares to the model. Storing the new information in a database creates the foundation for a closed control loop and further predictions.
By feeding information from the soft sensor back into the system, it can be available for other users and other plants – creating more reliable forecasting models to complement a golden batch monitoring trajectory.
The models in SIMCA® make it easy for your instrument users to get invaluable information from their Raman spectroscopy data. This includes formulas that help predict and control common problems to:
- Gain better control over yield
- Reduce batch failures
- Predict outcome for batches
- Lower production costs
MVDA is essential in all the phases of the production process development: discovery (to understand), manufacturing (to build a process), and quality assurance (testing).
Why Embed Data Analytics in Raman Instruments?
There are a number of good reasons for embedding SIMCA® data analytics from Sartorius into your spectroscopy or Raman instruments.
- Saves time and cost. Perhaps the most important reason to use SIMCA®-Q for embedded MVDA is that it saves time and costs in developing models that can be validated. Sartorius uses proven methods that come with validation, regulatory backing, and documentation.
- Supports automation. Reduce the risk of human error. Simplify and speed up routine multivariate analysis work for your customers. This is relevant both in terms of scaling issues and also having the same prediction models across tools.
- Protects your IP. Give your customers access to your data models without having to disclose your formulas. For example, you can help customers predict glucose or lactate using your models – giving them a head start while protecting your IP.
- Provides consistency. Using a standard for model building, such as SIMCA®, across all product lines and instruments, makes it easier to ensure consistent results and adoption.
- Simplifies data analytics. SIMCA® is the gold standard for MVDA in the pharma industry, with regulatory backing and documentation for models. You gain transparency of data analytics methodologies while simplifying data analytics for your customers.
- Enables customization. Customize the system to your own requirements. Sartorius can provide you with a sample code, help you implement it, and advise you on the customer perspective.
Spectroscopy Optimized Tools in SIMCA®
The latest version of SIMCA® has features that are especially useful for analyzing spectral data – making multivariate calibration easy and reliable. These include:
- Spectroscopy-specific workflow. Users can set or generate a spectral ID at import and select from an extended library of preprocessing algorithms.
- Preprocessing wizard. The preprocessing wizard provides an interactive, graphical wizard for filter settings, as well as settings for Python pre-processing.
- Calibration wizard. The calibration wizard visualizes model quality after each preprocessing option, which helps users choose the best prediction model.
Want to know more?
Watch the Sartorius Embedded Days webinar videos for Raman applications.
You’ll see case study examples from both Kaiser and Tornado Raman manufacturers.