Figure 1: Preprocessed bioreactor data collected from three different units.
by Kevin Broadbelt, Global Applications Scientist – Biopharma, Thermo Fisher Scientific
Biopharmaceutical manufacturing processes depend on living organisms to generate the product, making them extremely complex and susceptible to even slight variations in the bioreactor environment. Any changes in conditions can significantly impact yield and quality, with costly knock-on effects that include failed batches, inefficient use of resources and end products that don’t meet quality specifications. As a result, it is crucial that every aspect of the reaction conditions is fully understood to ensure optimal process control and consistent end-product quality. Biologics manufacture therefore depends on rigorous process development to optimize bioproduction workflows.
In-line and at-line monitoring of materials plays a vital role in this process and can often be streamlined by adopting real-time process analytical technologies (PATs), such as Raman spectroscopy. This technique allows early identification of issues that could lead to costly batch failures, enabling preventative action to be taken and improving efficiency. To date, process Raman as a PAT tool has largely been employed for upstream cell cultivation and harvesting in bioreactors, but it also has potential for downstream extraction of drug substance, and purification processes.
An introduction to Raman
Raman spectroscopy is a common analytical technique that allows researchers and manufacturers to define the chemical composition of solid, liquid or gaseous materials. The technique involves using a fiber-optic cable to direct laser light at a sample. The energy from the laser causes covalently bonded molecules in the sample to vibrate and scatter the light; this can be either elastic scattering – with the energy of the molecule unchanged after interaction with the photon – or inelastic scattering, where the molecule absorbs some of the energy and the scattered photon loses energy. The inelastically scattered light is collected and interpreted by a detector, generating a Raman spectrum that is unique to each molecule. This molecular fingerprint enables both qualitative identification of a given substance, and quantification of the amount of the analyte of interest present.
The combination of qualitative and quantitative data makes Raman spectroscopy one of the most powerful analytical technologies available for process monitoring, where it is known as process Raman. As well as high specificity, it offers rapid, non-destructive, and versatile sample analysis. Because it is non-destructive, it is ideal for integration directly into a production line for continuous in-line or on-line process monitoring. It is also fast, meaning that most substances can be measured in a matter of seconds.
Until recently, the main drawback of this technique was that it required complex, bulky, and expensive equipment, as well as specialist know-how to operate and maintain the instrument. However, this has changed with the introduction of compact, easy-to-use, reliable, and affordable systems – such as the Thermo Scientific Ramina Process Analyzer. The availability of smaller, portable devices with a simplified user interface enables biopharmaceutical manufacturers to easily integrate Raman spectroscopy into production processes, helping to improve efficiency and quality while making the most of valuable bench space.
The Raman signal produced in a bioreactor is complex due to the diverse mix of components within the system, and must be interpreted from its raw form into actionable metrics, like concentrations and substance identities. This relies on chemometrics – a subset of machine learning and artificial intelligence specifically tailored toward chemical analysis – which can develop models to interpret the information delivered by process Raman systems, allowing actionable insights for process development and feedback control.
Raman spectroscopy in biopharma process development
Raman spectroscopy can be used to measure solids, liquids, gases, powders or slurries, and this flexibility allows real-time monitoring at various points in the production process. It is particularly useful when measuring in a bioreactor – which may contain many different types of molecules – as Raman measurements are unaffected by water, so reactions in aqueous solutions to be monitored effectively. The direct, linear relationship between the concentration of a given substance and the intensity of the peaks in the Raman spectrum makes it easier to build quantitative models that accurately predict the concentration across the range of detection, even with a relatively small sample set. Raman spectroscopy can help to answer questions such as:
- Do the cells have the right amount of glucose?
- Are too many secondary metabolites building up?
- Are the cells beginning to produce the product of interest?
- How much product has been produced, and does it have the right characteristics.
Crucially, it does this in real time, enabling process adjustments as and when necessary.
A good example of the benefits of using process Raman in biopharmaceutical production is modeling of the glucose concentrations during a bioprocess. A glucose feeding cycle is often essential in biomanufacturing to enable cell reproduction. Modeling the glucose content can help to determine exactly when to add more glucose, ensuring a precise rate of cell production. This modeling begins with the collection of data. Figure 1 shows three Raman data sets collected from a bioprocess performed at varying global locations, using the same instrument set-up at each site.1 Despite the small data set, it was possible to quickly build an accurate and precise predictive global glucose model, using data science tools to clean (pretreat), combine and process the information. The key to the success of this approach lies in the stability, accuracy, and consistency of the analyzers, ensuring that every system delivers comparable results no matter where it is deployed. This enables the use of fundamental preprocessing methods that target and amplify the relevant signals within the Raman data.
Once established, the predictive model can be transferred to any other site using the same set-up, and the data outputs used for real-time tracking of the glucose concentration during the bioprocess. Figure 2 shows how the model derived from the data above has been applied to a fourth identical bioprocess and Raman system set-up. The glucose level steadily declines over time until a specified minimum value is reached. At this point, glucose is fed into the bioprocess and the level rises again, ensuring optimal process control.
Future applications in biopharma
Raman spectroscopy is already recognized as a valuable tool in upstream bioprocess development for protein-based biotherapeutics, offering effective real-time in-line monitoring that allows rapid correction of any variations in reactor conditions. However, there is currently a disconnect between upstream and downstream monitoring. During downstream extraction and purification, HPLC UV-VIS detectors are currently typically used to monitor the biologic of interest. However, this cannot be performed in real-time, and the results are prone to specificity issues, due to limitations in the detectors. Raman spectroscopy – with the correct validation – could offer a more specific solution, satisfying a clear need within the industry for a real-time PAT tool for downstream processes. Raman spectroscopy also has further potential across the full scope of biopharmaceutical production – including nucleic acids, and cell and gene therapies – with new Raman spectroscopy modalities under development for use in the production of these newer biotherapeutics.
Raman spectroscopy is increasingly being recognized as a useful tool for process development and monitoring in the manufacture of protein-based therapeutics, offering non-destructive compositional measurements for a variety of sample types. The evolution of smaller, less complex, hand-held systems means that the technique no longer requires a Raman specialist, and the small size and portability of these instruments offer flexibility in terms of placement and scalability. The real-time, actionable data provided by Raman analysis can help biopharmaceutical manufacturers to optimize and manage their bioprocesses for maximum efficiency, even for non-protein therapeutics.
References:
1. Application note. Using process Raman and data science to gain actionable information for glucose monitoring. AN-RAMINA-0522 v01. Thermo Fisher Scientific.