Contract development and manufacturing organizations (CDMOs) are driven to adopt new production technology because they need to cater to pharmaceutical firms that are interested in focusing on research and marketing. And today, pharmaceutical firms are, if anything, more interested in outsourcing production tasks than they ever were, given the increasing number of small- and medium-sized firms that lack manufacturing capabilities of their own.
It may sound as though CDMOs are spoiled for opportunity. However, CDMOs are anxious to distinguish themselves from their competitors and to offer multiple services to prospective customers, who often prefer to deal with as few service providers as possible. Both these considerations heighten the interest CDMOs have in acquiring novel capabilities.
At present, the capability most sought after by CDMOs is data-driven manufacturing, that is, the collection and processing of real-time data at each processing step to make manufacturing more efficient overall. Data-driven manufacturing elements include sensor technologies, advanced analytics, robotics, and even “digital twins,” virtual versions of physical systems that may be used to optimize processing.
Such elements are central to the Industry 4.0 approach, which involves the use of “smart” and autonomous systems to enhance the computer control of manufacturing. The Industry 4.0 approach is less common in the pharmaceutical industry than in industries with more consistent data formats and less demanding regulatory environments. Nonetheless, CDMOs are highly motivated to adopt the Industry 4.0 approach in the production of high-quality medicines such as biopharmaceuticals.
Delivering consistent quality
“Delivering a product with the right quality attributes consistently is critical to ensuring customer satisfaction for any CDMO,” says Thaddaeus Webster, PhD, a lead scientist at Lonza. “The utilization of innovative bioprocessing technologies increases the level of insight and control during the manufacture of biologics and biotherapeutics.”
He cites the move to utilizing inline spectral methods for process control as well as the use of multivariate analysis (MVA) to monitor and identify processes trending away from the norm as examples of how digitalization is being used to meet customers’ quality demands.
“Ultimately, reduced process variability leads to right-first-time delivery,” Webster notes, “potentially removing costly deviations, product delays, and customer dissatisfaction.”
Making a competitive case
For Atul Mohindra, PhD, the senior director of R&D at Lonza, offering innovative manufacturing technologies is an important part of winning business—but only a part. And like the other experts interviewed for this article, he stresses the importance of customer satisfaction.
“[Because we are] a CDMO that can provide an enhanced level of process control, our customers can have confidence that we will provide both drug substance and product at the desired quality and do so consistently,” Mohindra asserts. “However, having such technologies in place is really secondary to how the information generated is used.”
Sébastien Ribault, PhD, head of end-to-end solutions at MilliporeSigma, holds a similar view: “While our innovative technology offering is certainly a contributing factor, I believe we win business because we are using the right technologies at the right place. In addition to the advanced and innovative bioprocessing technologies we offer, we provide regulatory expertise and ease of use to help streamline the process for our customers, helping them get their therapies to market faster.”
Ribault adds that understanding how and when to use new production methods and technologies is key to taking full competitive advantage. “For example, you may use rapid sterility testing in the release process to speed up the drug release, but if you don’t have rapid virus testing, the time you’re saving on one side hits a bottleneck somewhere else,” he explains. “Successful CDMOs know how to best utilize technologies to innovate throughout the entire drug development process, rather than solely focusing on one step.”
Meeting data regulations
Regulatory demands for more information about how drugs are made is also impacting what drug firms expect of CDMOs. “There is no doubt that innovative bioprocessing techniques and technologies provide a competitive edge in our sector,” says Jadranka Koehn, PhD, senior director of innovation partnerships at Rentschler Biopharma. “This new dimension of digital maturity and capability is becoming increasingly important in the CDMO selection process, and its relevance extends beyond pure bioprocessing technologies.
“Our strategic vision drives us not only to offer exclusive services, but also to make digital differentiating capabilities available to our clients. This trend is reflected in the new FDA regulations, where data integrity and process control are two important cornerstones. Regulatory authorities expect us to understand our process and product in its entirety, hence the need for data, especially high-quality data.”
Koehn cites digital innovations such as mixed reality (the visualization of real and virtual objects), natural language processing, and context-sensitive human–machine interfaces as examples of technologies helping the CDMO sector to more fully model and control manufacturing processes.
Understanding how best to implement a digital manufacturing operation is an increasingly important skill for a CDMO. Part of the challenge is that industry experience of switching to a data-driven Industry 4.0 model is limited.
The cost and validation requirements differ between traditional and Industry 4.0 approaches, maintains MilliporeSigma’s Sébastien Ribault. “More traditional techniques will benefit from an existing data package that will help reduce the number of data points necessary for a validation,” he elaborates. “The pain points are known and can better be anticipated.
“Additionally, discussions with the regulatory agencies will be easier with technologies that have been used for several years versus new and innovative ones. This being said, it is again a case-by-case discussion to understand how much it differs.”
Data management should also be a consideration for any CDMO setting up a digital manufacturing operation. “Innovative manufacturing processes often have implementation requirements that are more involved compared to traditional techniques and technologies,” Lonza’s Thad Webster observes. “For instance, predictive models that utilize spectral inputs present a large amount of data that need to be stored.
“For monitoring and control purposes, bringing in the appropriate quality assurance, regulatory, and other subject matter experts is critical to quickly identifying and addressing technology gaps. Early involvement with these groups should help to facilitate a smooth adoption of proposed technologies for innovative manufacturing processes.”
Return on investment
Another bioprocessing expert that sees appropriate data infrastructure as a key CDMO attribute is Peter Crane, PhD, corporate strategy manager at Synthace, the developer of a cloud-based bioprocessing software platform. “A data foundation is essential,” he says. “A data foundation or infrastructure is as important as the facility and equipment.” He adds that other important elements include the deployment strategy (whether a cloud- or edge-based platform is used), application programming interfaces, and security and service level agreements around uptime and user interfaces.
“There is a cost associated with all of this, and it may be significant if the facility is erring toward legacy,” he continues. “To move toward a more innovative way of working, we have to bridge development and manufacturing and rebuild the stack from the ground up to be modular, automation friendly, and data friendly—and to have a digital–physical infrastructure at its core.”
But the extra effort and investment are worthwhile, advises Wolfram Schulze, vice president of information systems and organization at Rentschler Biopharma. Although Schulze acknowledges that the validation steps followed for a traditional process differ from those for an innovative process, he insists that the result “has to be the same.”
“With more high-quality data available, robust processes can be developed that act as an enabler for early-stage development,” he continues. “Full data integration and modeling enable faster process development than is possible with traditional techniques and technologies. Ultimately, agility, automation, and an iterative incremental approach will not only drive a learning culture, but also provide a safety net of semiautomated processes. This in turn will facilitate transfer of knowledge from the software domain with each successive step, resulting in increased quality.”
For Sébastien Ribault from MilliporeSigma, the benefits of digital technology for process monitoring are clear: “One example of how it applies to GMP manufacturing is the anticipation of a lack of oxygen for better process regulation. We have developed an algorithm that identifies the conditions under which the cell growth requires a specific oxygenation adjustment.
“With highly demanding cells, this adjustment can require human intervention. Our algorithm indicates to us two hours before the need for regulation that we will have to modify our settings. In the past, the regulation was linked to the event. Now, the regulation is linked to the event prediction.”
Lonza has also found ways to enhance its capabilities through investments in digital manufacturing. For example, investments in process analytical technologies (PAT) have allowed the company to upgrade its monitoring and control of critical process parameters in cGMP-compliant processes.
“We have already implemented several PAT sensors within the Lonza network, and we are currently investing even more in innovation within this area,” notes Atul Mohindra. “Furthermore, we are also investing in bioinformatics with the aim of implementing machine learning capabilities very soon.
“Gathering and using process data is at the core of Lonza’s manufacturing services. We have a large number of molecules. So, our clinical platforms allow us to gather huge bodies of data, to establish robust processes, and to ensure that our customers are successful at each critical milestone. In addition, the data facilitates the transfer of processes within our network—for example, from Slough to Hayward or from Slough to Singapore—or from our network to our customers’ facilities.”
In time, all CDMOs are likely to go digital, comments Synthace’s Peter Crane. He believes that customer demand for high-quality products and information will cause data-driven manufacturing to become the norm.
“Industry 4.0 isn’t a differentiator per se, but it’s a lever that allows an organization to deliver high-quality products more quickly, even with more complex biological systems,” he points out. “The buzzwords may get people excited at conferences, but in this industry, it’s all about being the trusted CDMO partner and delivering the best results to clients/patients.
“I expect that in the short term, companies will use these capabilities to differentiate themselves on the basis of the results the capabilities can provide. As more CDMOs digitalize, these capabilities will become commonplace—a ‘must have,’ not a ‘nice to have’—for many of these organizations.”
AI to Screen, Characterize, and Optimize Drug and Vaccine Candidates
TeselaGen Biotechnology has developed a modular AI-driven software platform for designing, building, and optimizing biological products, according to CEO and co-founder Eduardo Abeliuk, PhD. TeselaGen’s Discover module implements different machine learning and Bayesian optimization techniques to efficiently evolve genetic constructs in search of an optimal design, he explains, adding that the company has applied an arsenal of bioinformatic tools to achieve its goals.
In a recent study that involved a collaboration between TeselaGen and the Novo Nordisk Foundation Center for Biosustainability, it was demonstrated that artificial intelligence could be used in combination with mechanistic models to optimize the titer of a bioproduct whose biosynthesis is controlled by a highly regulated pathway.
“In this study, a nearly two-fold increase in titer was achieved within a single experimental iteration,” reports Abeliuk. “We have made these advanced tools readily available for other applications in bioprocessing, where pharmaceutical companies, for example, might be interested in optimizing monoclonal antibody production. In this context, these computational techniques could be used not only to increase titer, but also to optimize other properties of interest, such as cell growth, anti-apoptosis, or the ability to grow under certain defined media.”