AI transforms cell & gene therapy manufacturing with real-time insights

The application of artificial intelligence (AI) in manufacturing processes is changing how companies in the life sciences industry monitor and control production, particularly in the area of cell and gene therapy (CGT).

Real-time batch monitoring

Machine learning and digital twin technology are being used to improve process control in real time, allowing for the early detection of deviations and enabling predictive adjustments. This approach is intended to safeguard product quality and reduce the risk of production failures and wasted resources.

Smriti Khera, Head of Global Life Sciences Strategy and Marketing at Rockwell Automation, said: "Real-time batch monitoring powered by machine learning is helping manufacturers detect deviations early to reduce failures, improve consistency, and cut costs. By simulating cell and gene therapy processes with real-time data, digital twins enable risk-free optimisation of culture conditions, purification steps, and nutrient strategies, all of which can be tailored to each patient's unique cells. AI and predictive analytics support FDA's Quality by Design by building quality into the process and accelerating access to therapies."

High costs and failure risks

The costs associated with CGT manufacturing are substantial, with a single batch possibly exceeding USD $500,000. Failures arising from minor variations in the production process not only elevate costs but also delay patient treatments. This environment is driving manufacturers to adopt more sophisticated monitoring solutions as the demand for these therapies grows globally and more treatments move from clinical trials to commercial production.

"It's no secret that cell and gene therapy (CGT) manufacturing is an expensive process - a single cell therapy batch can cost upwards of $500,000 to produce. When a batch fails due to minor process variations, both manufacturers and patients pay the price in dollars and treatment delays," Khera noted.

With more than 2,000 clinical trials worldwide and regulatory bodies such as the US Food and Drug Administration (FDA) approving multiple therapies, the industry faces pressure to scale production while maintaining stringent quality and cost controls.

Process variability challenges

CGT manufacturing is particularly sensitive to process changes due to the use of living cells, which can react strongly to shifts in their environment. Khera explained: "Even minor variations can have outsized consequences in CGT manufacturing. Unlike conventional drugs, these therapies use living cells that respond dramatically to subtle environmental shifts. Minor temperature fluctuations of just 1-2°C can trigger cellular stress responses. Slight changes in media composition can also affect growth rates. Inconsistent centrifugation between batches can impact cell viability. These deviations compound throughout the process, compromising therapy potency and safety."

Historically, quality control has depended on testing after production was completed. However, Khera stressed that these checks often come too late to save compromised product: "Traditional quality control methods, such as testing after production, usually come too late, with the damage to the batch already done. However, AI and predictive analytics are reforming CGT manufacturing by catching problems before they ruin a batch."

AI in manufacturing

Advanced AI tools are now taking a more proactive approach to quality assurance. "AI-powered batch monitoring systems function as tireless quality inspectors, simultaneously analysing thousands of process parameters to catch subtle patterns that human operators could miss. These systems harness several AI technologies: Computer vision examines bioreactor imagery to assess cell characteristics. Machine learning processes sensor data tracking pH, oxygen, glucose, and metabolites. Natural language processing scrutinises batch records to find connections between procedural variations and outcomes," Khera said.

"AI's true power lies in detecting deviations before they affect product quality. For example, when monitoring cell cultures, AI can identify early metabolite concentration shifts that signal potential future problems, allowing operators to make real-time adjustments to temperature, pH, and nutrients."

By leveraging these technological advancements, manufacturers can expect greater consistency, fewer failures, and less waste. According to Khera: "With the help of AI-powered batch monitoring systems, manufacturers get fewer batch failures, more consistent quality, and less waste. These systems go beyond providing basic alarms, they can be trained to understand how various factors impact each, enabling them to distinguish between normal variations and real anomalies that need intervention."

Implementation strategies

Khera advised that organisations adopt a staged approach when deploying AI and digital twins, beginning with high-return-on-investment projects. "Successfully implementing AI and digital twins in CGT typically requires a staged approach, starting with high-ROI use cases. Everything begins with data quality - ensuring sensors and collection systems deliver accurate measurements. This often requires equipment upgrades and calibration protocols," she said.

Teams with expertise across engineering, quality control, and data science are important for turning data into effective decision-making tools. "Cross-functional teams comprised of engineers, quality specialists and data scientists are crucial for creating actionable insights. Shrewd organisations start with historical batch analysis before attempting real-time monitoring, building confidence as they go," Khera added.

This allows companies to build up confidence and value as implementation progresses to more advanced uses of AI.

From production to patient

Improving manufacturing processes is seen as essential for increasing patient access to advanced therapies and for meeting regulatory expectations. Khera stated: "As cell and gene therapies break into mainstream medicine, making manufacturing more efficient becomes key to getting these treatments to more patients at prices they can afford. AI and predictive analytics align with the FDA's Quality by Design approach, which involves building quality into the process from the outset instead of simply testing at the end. With AI providing unprecedented insight into manufacturing, companies can show regulators exactly how they maintain control."

Machine learning and automation are anticipated to help standardise and improve medicine manufacturing over time. "As systems get smarter, we're seeing manufacturing that learns and improves on its own. These intelligent systems don't just maintain quality - they improve it, bringing science closer to advanced therapies that work dependably in both production and patients," Khera concluded.