One-Batch Calibration: A Process-, Instrument-, Scale- & Site-agnostic Method For Streamlining Raman Implementation
Raman spectroscopy is a process analytical technology (PAT) tool used in both small and large molecule pharmaceutical processes. It offers numerous opportunities to improve process understanding and optimize product quality and process efficiency, building quality into the biopharmaceutical process. In the biopharma sector, the capability of Raman spectroscopy to monitor cell cultures in real time and with high molecular specificity makes it an excellent tool for both upstream and downstream processes.
Read more about
Raman Implementation: Limitations of Traditional Approaches
Raman spectroscopy enables the user to monitor critical process parameters in-line by simply inserting a probe into the process – an efficient, non-destructive approach – and in real time. This technology is particularly valuable in a biopharmaceutical process, where it serves to monitor and control key parameters such as glycosylation, viable cell density, glucose and amino acid feeds.
However, prior to its use in the process, implementation and calibration is a critical step. Traditional methods to Raman implementation rely on time-consuming and resource-intensive processes including the measurement of several batches to generate a sufficient amount of spectral data. These are needed to build robust and reliable calibration models that correlate Raman spectra with the off-line reference measurements of target parameters. With this traditional, multi-batch approach, the model-building phase can take up to several months, resulting in significant costs. As a result, Raman implementation in the biopharmaceutical industry today is rather slow, despite the numerous advantages it offers once successfully integrated into the process.
Accelerating Raman Implementation for Real-Time Cell Culture Monitoring
An innovative approach has been identified which overcomes limitations of traditional, time-consuming Raman implementation methods, eliminating the need for multiple calibration batches: the one-batch calibration approach.
Key benefits include:
- A streamlined workflow: Simplified Raman implementation process which saves time, costs, and resources.
- Unmatched flexibility: Seamless adaptation to the broad variety of bioprocess conditions, including different cell-lines, clones, media, and feeds, as well as facilitation of site-to-site technology transfer and scale-up.
- Enhanced robustness: Improved reliability of Raman measurements through advanced algorithms and statistical techniques.
The direct comparison of standard multi-batch calibration and one-batch calibration methods clearly shows the numerous benefits of the streamlined approach: reduced number of batches, lower sampling and measurement efforts, less resources, shorter implementation times, and higher robustness and flexibility with regard to media, feed, cell line, clone, or instrumental changes. An additional benefit is the efficient technology transfer from one site to the other or to a different scale. At the same time, one-batch calibration provides equivalent monitoring accuracy as is achieved using traditional methods.
Benchmarking Study: A Comparison of Standard and One-Batch Calibration Methods
To evaluate the suitability of the novel one-batch calibration approach for Raman implementation, a benchmarking study was performed comparing the accuracy of one-batch and multi-batch calibration methods.
Materials and Methods
Fed-batch cell cultures (n = 5) were performed using a 250 mL modular bioreactor with a CHOZN® GS -/- cell line producing IgG1. An EX-CELL® Advanced CHO fed-batch medium was used with the following set process parameters: 37 °C, pH 7, 40% dissolved oxygen (DO) and 300 rpm agitation speed. In the benchmark study, an Ambr® 250 mL modular bioreactor (Sartorius AG, Goettingen, Germany) was used, however, the approach is also compatible with other 250 mL bioreactors such as the DASbox® Mini Bioreactor System (Eppendorf SE, Hamburg, Germany) or the Applikon MiniBio (Getinge AB, Goeteborg, Sweden) and with larger scales bioreactors.
Samples were taken twice a day during batch runs and subjected to the following measurements: Concentrations of glucose, lactate, and viable cell density (VCD) using an automated cell culture analyzer; mAb titer measurements using two-dimensional liquid chromatography coupled to tandem-mass spectrometry (2D-LCMS); and Raman spectroscopy using the ProCellics™ Raman Analyzer. Bio4C® PAT Raman Software was used for data acquisitions and Bio4C® PAT Chemometric Expert for data analysis.
Five batches were run in total. For the standard calibration method, data was collected from the first three batches (batch 1-3). For the one-batch calibration method, data was collected from the first batch only (batch 1). The parameters used to evaluate the model calibration performance were the explained variance coefficient R², the cross-validated correlation coefficient Q², and the root mean square error of cross-validation (RMSECV).
After calibration, both models were used to monitor batches 4 and 5 (“monitoring batch 1 and 2”). Performance and accuracy were compared by measuring glucose, lactate, VCD, and mAb titer over a process time of 12 days and calculating root mean square error of prediction (RMSEP) and relative error of both methods. The calculation of relative error values (defined as the ratio between RMSEP and the maximum value of the range in the validation batches) for each parameter allows for comparability of results across different units.
Evaluation of Model Calibration Performance
Table 2 confirms that despite a significantly different number of batches and samples – 3 vs. 1 batches and 40 vs. 13 samples for standard and one-batch methods, respectively –calibration performance of both methods is comparable. The performance of the streamlined one-batch calibration is even superior to the standard, multiple-batch approach. This has been validated by R2 and Q² values close to the ideal value of 1 as indicators of a high-performance model both in calibration and cross-validation, and RMSECV with low values, collectively attesting to the model's robustness, accuracy, and precision.
Comparison of Method Reliability and Prediction Accuracy
Figure 1 shows Raman plots of monitoring batches 1 and 2 generated based on one-batch calibration, standard calibration, and off-line reference measurements of glucose, lactate, viable cell density, and titer. The plots highlight that measurements based on one-batch calibration method can track the process kinetics with uncompromised accuracy despite a significantly reduced effort needed for implementation.
Monitoring Batch 1 Results
Monitoring Batch 2 Results
Figure 1: Plot of Raman predictions comparing one-batch calibration method, standard calibration method, and off-line reference measurements for monitoring batches 1 and 2.
Values shown in Table 3 confirm the robustness of a model based on the one-batch calibration method in comparison to the standard, multiple-batch approach. While both methods yield comparable results for VCD and titer, one-batch calibration stands out for its significantly improved accuracy in glucose and lactate measurements, as evidenced by a reduction in relative error values from 7% to 3% for glucose and from 10% to 3% for lactate. This represents a two-fold improvement for both parameters compared to the standard calibration method.
Figure 2 visualizes the accuracy of both methods: Performance in terms of accuracy is satisfactory for all measured parameters for both one-batch and standard calibration methods. The significantly improved accuracy for glucose and lactate measurements based on the one-batch calibration method indicates excellent performance, good reliability, and strong predictive ability of this approach.
Figure 2:Average relative errors for monitoring batches 1 and 2, comparing the performance of standard and one-batch calibration methods.
Empowering Raman Users with Streamlined Calibration
The benchmarking study confirmed the suitability of the innovative one-batch calibration method for the development of a robust and accurate model, eliminating the need for multiple batches and long calibration runs. Results were comparable, in several cases significantly improved with the one-batch calibration method in comparison to standard, multiple-batch calibration. Study results show how the one-batch calibration method overcomes limitations of traditional model building methods and provides an alternative approach that is more efficient, accurate, and flexible.
The one-batch calibration method is a groundbreaking approach which empowers users to navigate their Raman implementation process with confidence. It is a streamlined strategy enabling a seamless adaptation of Raman technology to various cell-lines, clones, media, bioreactor scale, feed conditions, instruments, and facilitating easier site-to-site technology transfer.
By leveraging innovative algorithms and statistical techniques, coupled with a proprietary model library, the accuracy and reliability of Raman modeling are enhanced, enabling users to make confident decisions backed by robust data. This way, the usual laborious and complex processes associated with Raman calibration can be replaced by a streamlined workflow that saves time, effort, and valuable resources.
Para seguir leyendo, inicie sesión o cree una cuenta.
¿No tiene una cuenta?