Damien O’Connor, Associate Director, Cognizant Life Sciences Manufacturing05.03.23
Because manufacturing quality has everything to do with the safety and efficacy of the final bioproduct, manufacturers are continually looking for ways to gain more insight into processes. Increasingly, pharma’s manufacturing engineers are finding the variation control and efficiency they need by applying process analytical technologies (PAT) to acquire actionable data directly from the process in real time.
Although the FDA’s objective of achieving significant health and economic benefits by “application of modern process control and tests in pharmaceutical manufacturing” still stands, the up-take of PAT by pharma, in general, has been slow, a conservative pace even more pronounced in the biotech space.1
Fortunately, adoption is accelerating as the industry transforms operations digitally to Pharma and Biopharma 4.0 models post-pandemic. PAT data can become extremely useful and valuable creating stable, reliable, repeatable bioprocesses. It is also valuable because it demonstrates essential process quality control to regulators.
Recent projections from P&S Intelligence value the global market for PAT approaching $14 billion by 2030. According to the firm’s analysis, the pandemic indeed has been a strong driver of market growth, accelerating the adoption of PAT to monitor and control production and processes by pharmaceutical, biopharmaceutical and contract research organizations. Because PAT is primarily used to monitor and evaluate the drug development process, the demand for the technology soared in 2020.2
Sensing technologies for PAT application are also under constant development but, currently, process engineers are relying on spectroscopy (molecular, atomic and mass), chromatography (liquid and gas), capillary electrophoresis and particle size analysis to create a holistic under-standing of process conditions.
A single PAT measurement is a spectrum consisting of a list or arrays (time, channel, value) that cannot be stored in a classical data historian. It is a distinct pain point and an issue that is constraining the successful uptake and utilization of the technology by the biotech industry (see Figure 1).
Figure 1. Several spectra forming a multi-dimensional time series.
One way the biotech process engineers are overcoming these issues is to enlist the help of ex-pert solutions providers with the experience and software platforms to integrate PAT data, regardless of source, into the organization’s existing biotech data infrastructure.
The solution is to extend the existing historian with databases that allow a more flexible schema. This results in better utilization of existing equipment and data that enables context-specific analysis.
Spectra are often stored in SQL-type databases as plain tables, separate from the other manufacturing data stored in the historian. The main problem with this approach is the loss of equipment and batch context. This is problematic to any subsequent analysis.
So why are spectra stored separately? Because most industrial data historian store values as simple time series in different data types, typically bool, int, float and string. Each time series point is a tuple of a timestamp and a single value (scalar). For PAT and other use cases, it would be required to extend the existing data shapes to accommodate vectors, matrices, and tensors (see Figure 2).
Figure 2
There are many use cases for these data structures:
Regular Time Series of scalars
• Temperature
• Pressure
• Level
Time Series Vector
• PAT spectra
• Vibrational spectroscopy for predictive maintenance
Time Series Matrix
• PAT
• Camera systems
Time Series Tensors
• Machine Learning or Deep Learning Models
• Machine Learning or Deep Learning Predictions
Several database platforms offer time series data handling utilities such as AVEVA, aspentech, and InfluxDB. Successfully extending the historian requires deploying a new source and linking it to a time series database that supports time-based vectors, matrices, and tensors. An example of this would be where Cognizant specialists have leveraged the AVEVA asset framework, to provide a feature rich environment to contextualize data (see Figure 3).
Figure 3. Extending the historian with a database platform approach.
The RAMAN spectra are attributes on the unit/vessel or located on the RAMAN equipment. Therefore, extending the existing AVEVA AF data model allows the measurements to be analyzed in the present batch or event-frame context.
PAT process monitoring data also plays a significant role in establishing critical process parameters (CPPs), which along with CQAs underpin quality by design (QBD) process development. That’s a dense alphabet, but key to accomplishing the goals laid out by regulators for their push to implement PAT in advancing drug quality, safety, and efficacy.3
For certain process applications the industry is moving towards closed loop control e.g. nutrient feeds in bioreactors, thus removing the need for manual sampling and feeds. We only see further efforts to take advantage on PAT to remove manual interventions.
In the meantime, leveraging multivariate analytics is proving an effective strategy to help im-prove process monitoring and control and establish the quality and control parameters for all process stages and events. For biotechs, especially those developing large-molecule drugs to fight prominent chronic diseases, the process is the product. An efficient cost-effective process is an important key to any drug product’s commercial success.
Establishing the most efficient process is a development and commercialization priority for biotech developers. Multivariate analytics plays a significant role in shaping a given process’s automation and control strategy, to calculate process envelopes such as “Golden Batch” profiles.
His treatise is straightforward. In batch processing, there are myriad sources of variation - from subtle variations in raw materials to shifting environmental conditions and fluctuating plant utilities. The point he makes is this: simply automating the process to duplicate the golden recipe won’t necessarily be the ticket to duplicating the recipe exactly the same, batch after batch. The best approach he notes, is to use data gathered from across the process then use multivariate analysis to shape controls and automate the process to combat variation. The result will be a process design that sustains a “Golden Zone” from which product yield and quality can consistently emerge batch after batch.4
Once integrated into existing data infrastructure, PAT data can become a powerful and significant, singular source of process “Truth,” the kind that reduces risk and sustains the development and commercialization timelines of today’s most effective biotech therapeutics. The data from in-line, on-line and at-line sensors can be leveraged in a variety of ways to achieve process and product quality goals. Even though the data collected by at-line and off-line sensors and analytical systems can yield actionable information, the insights the data offers may come too late to manage emerging variation in the batch currently under process.
PAT’s uptake by the industry to monitor and control commercial-scale biotech processes is accelerating, but its value can’t be realized if the data can’t be disseminated and analyzed properly. Existing and legacy data systems as well as stable, validated processes and the extremely conservative, regulated character of the industry all contribute to governing the pace of technical change. Fortunately, the IT community is helping biopharma’s top innovators step up their pace with solutions that help overcome barriers to entry and deliver everything processors need to know now to gain ultimate control of bioprocesses.
References
1. https://pubmed.ncbi.nlm.nih.gov/20480150/
2. https://www.prnewswire.com/news-releases/process-analytical-technology-market-to-surpass-13-626-5-million-revenue-by-2030--says-ps-intelligence-301444977.html
3. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8234957/
4. https://www.linkedin.com/pulse/fallacy-golden-batch-cosimo-caforio/
Damien O’Connor is Associate Director at Cognizant Life Sciences Manufacturing.
Although the FDA’s objective of achieving significant health and economic benefits by “application of modern process control and tests in pharmaceutical manufacturing” still stands, the up-take of PAT by pharma, in general, has been slow, a conservative pace even more pronounced in the biotech space.1
Fortunately, adoption is accelerating as the industry transforms operations digitally to Pharma and Biopharma 4.0 models post-pandemic. PAT data can become extremely useful and valuable creating stable, reliable, repeatable bioprocesses. It is also valuable because it demonstrates essential process quality control to regulators.
Recent projections from P&S Intelligence value the global market for PAT approaching $14 billion by 2030. According to the firm’s analysis, the pandemic indeed has been a strong driver of market growth, accelerating the adoption of PAT to monitor and control production and processes by pharmaceutical, biopharmaceutical and contract research organizations. Because PAT is primarily used to monitor and evaluate the drug development process, the demand for the technology soared in 2020.2
Data integration is key to enabling process and operational efficiency
Most biotech and pharmaceutical companies are adopting PAT in manufacturing to provide re-al-time operational insights that allow better control and lead to higher yields, purity, and/or shorter cycle times. Biotech processors and drug product manufacturers are still developing innovative ways to apply PAT in process. Control engineers are installing in-line and on-line sensors and analytical systems to monitor process streams and conditions within bioreactors and vessels for continuous control and in-situ analysis.Sensing technologies for PAT application are also under constant development but, currently, process engineers are relying on spectroscopy (molecular, atomic and mass), chromatography (liquid and gas), capillary electrophoresis and particle size analysis to create a holistic under-standing of process conditions.
What we have is a failure to communicate
Introducing PAT into an existing production environment or process has distinct challenges. Although the implementation of in-line, on-line and at-line sensing can be relatively straightforward, the firehose of spectral data PAT systems generate can overwhelm existing biotech data infrastructures centralized through a legacy data historian.A single PAT measurement is a spectrum consisting of a list or arrays (time, channel, value) that cannot be stored in a classical data historian. It is a distinct pain point and an issue that is constraining the successful uptake and utilization of the technology by the biotech industry (see Figure 1).
Figure 1. Several spectra forming a multi-dimensional time series.
One way the biotech process engineers are overcoming these issues is to enlist the help of ex-pert solutions providers with the experience and software platforms to integrate PAT data, regardless of source, into the organization’s existing biotech data infrastructure.
Strategies for integrating PAT data effectively
Classical data historians have been developed for scalar time series information. This has worked for most sensor types and data, but data historians are becoming notorious for their inability to accommodate higher dimensional time series information.The solution is to extend the existing historian with databases that allow a more flexible schema. This results in better utilization of existing equipment and data that enables context-specific analysis.
Spectra are often stored in SQL-type databases as plain tables, separate from the other manufacturing data stored in the historian. The main problem with this approach is the loss of equipment and batch context. This is problematic to any subsequent analysis.
So why are spectra stored separately? Because most industrial data historian store values as simple time series in different data types, typically bool, int, float and string. Each time series point is a tuple of a timestamp and a single value (scalar). For PAT and other use cases, it would be required to extend the existing data shapes to accommodate vectors, matrices, and tensors (see Figure 2).
Figure 2
There are many use cases for these data structures:
Regular Time Series of scalars
• Temperature
• Pressure
• Level
Time Series Vector
• PAT spectra
• Vibrational spectroscopy for predictive maintenance
Time Series Matrix
• PAT
• Camera systems
Time Series Tensors
• Machine Learning or Deep Learning Models
• Machine Learning or Deep Learning Predictions
Take a software platform approach to data historian extension
One strategy proving effective for managing PAT datasets is to extend the existing historian with databases to manage the higher dimensional time series information the sensors are capable of delivering.Several database platforms offer time series data handling utilities such as AVEVA, aspentech, and InfluxDB. Successfully extending the historian requires deploying a new source and linking it to a time series database that supports time-based vectors, matrices, and tensors. An example of this would be where Cognizant specialists have leveraged the AVEVA asset framework, to provide a feature rich environment to contextualize data (see Figure 3).
Figure 3. Extending the historian with a database platform approach.
The RAMAN spectra are attributes on the unit/vessel or located on the RAMAN equipment. Therefore, extending the existing AVEVA AF data model allows the measurements to be analyzed in the present batch or event-frame context.
Creating more intelligent informed processes with PAT data
Once extended, the historian can display robust PAT spectral data (RAMAN) for a running batch. Cognizant’s specialists developed an add-in to AVEVA PI Vision to display raw PAT spectra and perform peak height calculations on the PAT spectra. This enables process monitoring of sensor data (temperature, pressure, pH) and PAT data side-by-side. This unique capability helps to deliver a tighter integration of PAT spectra in the manufacturing environment and a more effective means to integrate PAT spectra data to support the kind of multi-variate analytics needed to discover stable, validatable and efficient biotech processes.Leveraging PAT data for high-fidelity multivariate analytics
Based on monitoring and observing unit processes, engineers using PAT can determine critical quality attributes (CQAs) and intermediate quality attributes (IQAs) and help establish a continuous process verification (CPV) control strategy as well as real-time release testing (RTRT) protocols for their process.PAT process monitoring data also plays a significant role in establishing critical process parameters (CPPs), which along with CQAs underpin quality by design (QBD) process development. That’s a dense alphabet, but key to accomplishing the goals laid out by regulators for their push to implement PAT in advancing drug quality, safety, and efficacy.3
For certain process applications the industry is moving towards closed loop control e.g. nutrient feeds in bioreactors, thus removing the need for manual sampling and feeds. We only see further efforts to take advantage on PAT to remove manual interventions.
In the meantime, leveraging multivariate analytics is proving an effective strategy to help im-prove process monitoring and control and establish the quality and control parameters for all process stages and events. For biotechs, especially those developing large-molecule drugs to fight prominent chronic diseases, the process is the product. An efficient cost-effective process is an important key to any drug product’s commercial success.
Establishing the most efficient process is a development and commercialization priority for biotech developers. Multivariate analytics plays a significant role in shaping a given process’s automation and control strategy, to calculate process envelopes such as “Golden Batch” profiles.
A better recipe for fighting process variation with PAT data
The concept of the Golden Batch is straightforward. To be “Golden” a batch transits the process without issue, offers predicted yields and delivers rock-solid final product quality. But setting a Golden Batch recipe, posits John F. MacGregor, CEO of multivariate analysis consultancy ProSensus Inc., should not really be the end goal. He points out that once this dream batch has been created, automating to replicate this “sweet spot” may not actually serve in the process’ overall best interests.His treatise is straightforward. In batch processing, there are myriad sources of variation - from subtle variations in raw materials to shifting environmental conditions and fluctuating plant utilities. The point he makes is this: simply automating the process to duplicate the golden recipe won’t necessarily be the ticket to duplicating the recipe exactly the same, batch after batch. The best approach he notes, is to use data gathered from across the process then use multivariate analysis to shape controls and automate the process to combat variation. The result will be a process design that sustains a “Golden Zone” from which product yield and quality can consistently emerge batch after batch.4
Once integrated into existing data infrastructure, PAT data can become a powerful and significant, singular source of process “Truth,” the kind that reduces risk and sustains the development and commercialization timelines of today’s most effective biotech therapeutics. The data from in-line, on-line and at-line sensors can be leveraged in a variety of ways to achieve process and product quality goals. Even though the data collected by at-line and off-line sensors and analytical systems can yield actionable information, the insights the data offers may come too late to manage emerging variation in the batch currently under process.
Toward Biopharma’s 4.0 future
PAT’s sensing, monitoring and analytical systems are proving capable of delivering the robust data biopharma needs to support all key patient, product development and commercial objectives. That’s why PAT data has become so important to biotech development and central to Biopharma 4.0 operating models.PAT’s uptake by the industry to monitor and control commercial-scale biotech processes is accelerating, but its value can’t be realized if the data can’t be disseminated and analyzed properly. Existing and legacy data systems as well as stable, validated processes and the extremely conservative, regulated character of the industry all contribute to governing the pace of technical change. Fortunately, the IT community is helping biopharma’s top innovators step up their pace with solutions that help overcome barriers to entry and deliver everything processors need to know now to gain ultimate control of bioprocesses.
References
1. https://pubmed.ncbi.nlm.nih.gov/20480150/
2. https://www.prnewswire.com/news-releases/process-analytical-technology-market-to-surpass-13-626-5-million-revenue-by-2030--says-ps-intelligence-301444977.html
3. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8234957/
4. https://www.linkedin.com/pulse/fallacy-golden-batch-cosimo-caforio/
Damien O’Connor is Associate Director at Cognizant Life Sciences Manufacturing.
