In biopharmaceutical manufacturing, consistency is vital but hard to achieve. Turning raw materials into safe, high-quality medicines requires stable processes. However, the living cellular expression systems involved change over time, introducing variation.
Addressing this dichotomy is a focus for engineers with continuous process verification (CPV)—a risk-based method designed to ensure a process remains in a state of control throughout its lifecycle—being a common solution. CPV relies on real-time monitoring and analysis of critical parameters and quality attributes to confirm the process is producing drugs with the desired characteristics.
The challenge is understanding how best to use the large amount of process data gathered during CPV, says Mario Stassen, PharmD, a consultant at Stassen Pharmaconsult, who argues in a new paper that AI is a potential solution.
“Compared to traditional CPV approaches, which rely on predefined statistical models, AI introduces adaptability, learning from evolving datasets and improving process robustness over time. This is particularly valuable in biologics, where process variability is high, and subtle correlations between CPPs and CQAs may not be well understood using conventional methods,” he tells GEN.
“AI has the potential to revolutionize CPV by enhancing process understanding, anomaly detection, and predictive control. Traditional statistical process control (SPC) methods, such as control charts, often struggle with complex, multivariate relationships. AI models—particularly machine learning algorithms—can detect subtle, nonlinear trends that signal deviations before they escalate.”
Data comparisons
AI models can also compare historical and real-time data to predict optimal process conditions, enabling dynamic adjustments to maintain product quality, Strassen says, also citing the ability to make digital twins as a benefit.
“AI can create digital replicas of manufacturing processes, allowing companies to model ‘what-if’ scenarios, optimize conditions, and predict the impact of changes before implementing them in production.”
Adopting any new technology is complex, and AI is no different.
Strassen notes that “Biopharma companies must carefully consider the regulatory, data integrity, and infrastructure requirements to ensure successful implementation.”
Addressing regulators’ concerns will be particularly important, according to Strassen, who says, “While AI adoption in GMP settings is still evolving, regulatory agencies emphasize explainability, traceability, and validation of AI models.”
For example, AI is covered in the FDA’s Process Validation Guidance, EMA’s Annex 15 on Process Validation, and GMP Annex 11 on Computerized Systems as well as the ICH Q8–Q12 principles for continuous process control and lifecycle validation.
So, complying with regulators’ expectations will take work. However, the effort will be worth it, maintains Strassen, who says, “AI holds tremendous promise for enhancing process validation and control in biopharmaceutical manufacturing.
“By enabling real-time monitoring, predictive analytics, and adaptive process control, AI can significantly improve product consistency, regulatory compliance, and operational efficiency.”
