Why AI Advances in CGT Require Getting Data Right
Addressing the bottlenecks in cell and gene therapy manufacturing through a well-designed artificial intelligence approach is essential to avoiding data siloes.
The future of medicine is being shaped by cell and gene therapies (CGTs), which involve complex manufacturing steps. Because they are mostly personalized medicines today, production is limited to one patient dose at a time, and slowed by cumbersome and largely manual processes.
In addition to these inefficiencies, the field is still just learning which manufacturing parameters are most critical to patient outcomes, and how to optimize them.
As with so many areas, new tools such as artificial intelligence (AI) and machine learning are explored to optimize the steps to cut costs and increase efficiency while improving outcomes. However, doing so relies on access to reliable data, which has been a challenge.
Pharma’s penchant for data siloing is throttling the lifeblood of algorithms, leaving those working across the industry—particularly makers of enabling technologies—as key partners for CGT developers to ensure that crucial data can be leveraged in ways that allow AI to learn from it.
Understanding the bottlenecks
Since the first CAR T-cell therapy approval in 2017, the field of gene-modified cell therapies has progressed rapidly. Over the last seven years, FDA has approved 12 of these therapies, including the first T-cell receptor gene therapy for the treatment of metastatic synovial sarcoma in adults.1 Manufacturing processes account for a major portion of the cost of life-changing therapies, which has meant commercial access is so far limited to a few thousand patients per year, which could be a threat to continued progress.
The process starts with cell collection, followed by cell manipulation steps such as cell isolation, activation, modification, expansion, and final formulation. The final drug product is infused under the supervision of trained medical professionals. At each of these steps, there are opportunities to optimize.
Doing so starts from automating unit operations and enhancing digital data capture capabilities. Over the last decade, the CGT field has moved in this direction, and advanced automated unit operations are available. However, adopting these technologies in manufacturing could be faster.
The challenge begins with early-stage developers, who generally rely on manual processes with limited digital data capture. As the target drug candidate moves through development stages, manual data needs to be digitized, which may require more than easy transfer and integration among the manufacturing steps.
Identifying opportunities
Currently, the industry is focused on data capture in the cell collection, transport logistics, and therapy manufacturing processes. Often overlooked is the cell collection process, where apheresis is a commonly used technique and can be automated. Optimizing apheresis protocols has become a crucial part of clinical development for cell-based therapies, and the opportunities for data-driven AI are becoming more evident.
Researchers led by John Manis from Boston Children’s Hospital and Harvard Medical School demonstrated what is possible.2 The team performed a retrospective analysis of collection procedures for patients with sickle cell disease (SCD), performed on automated apheresis systems. The process is difficult for these patients, because SCD changes blood viscosity and causes components to clump. But by implementing new methods driven by the analysis of this data, the group observed that CD34+ collection efficiency increased from 4.9% to 36.8%—more cells in less time. The real-world applications became evident shortly after the results were presented, when regulators approved two autologous cell-based therapies for SCD.
The study is just one example of on-going data-based optimization efforts in cell collection. In the coming years, the industry would expect to see greater implementation of prediction algorithms and machine learning models that allow us to understand and adapt processes to improve efficiency, especially in cell collection methods.
Cell culture is another critical step in cell therapy production that may benefit from enhanced data collection, and could potentially lead to AI-based optimization. It is a tedious and complex process, and each subtype within the diverse set of cell therapy applications utilizes cell culture under a unique set of conditions.
The data collection and analysis for this step is centered on the utilization of biosensing tools in order to monitor the cell culture environment. In a recent study, researchers used such data to predict the optimal cell culture parameters using a machine learning model,3 achieving high cell densities in a stirred tank bioreactor while decreasing the use of process reagents, shrinking the cost of goods by 35% to 50%.
Similar opportunities are available with other automated platforms, which can be connected with biosensors to enable on-line monitoring of critical process parameters.4 Metabolic data such as lactate generation and glucose consumption during the cell culture could be used to predict the number of cells during the culture, and an appropriate time to harvest the cells. But researchers need to collect more data on these parameters than is currently available in order to train the prediction models and optimize cell culturing systems.
The fast-evolving landscape of CGTs must lean into the potential of AI and machine learning to revolutionize manufacturing processes. The importance of data collection throughout the value chain is evident, and it is a clear area where enabling technologies can improve understanding of the processes. As more processes are automated throughout the manufacturing chain, it is becoming critical to integrate different platforms and the data they can produce.
The focus on data-driven approaches to improve cell collection and culture processes underscores the ongoing efforts to enhance the production of impactful therapies. These advancements pave the way for a future where innovative treatments are more accessible and efficient, ultimately benefiting patients and the healthcare industry as a whole.
References
1. FDA Approves First Gene Therapy to Treat Adults with Metastatic Synovial Sarcoma. FDA. August 2, 2024. https://www.fda.gov/news-events/press-announcements/fda-approves-first-gene-therapy-treat-adults-metastatic-synovial-sarcoma
2. Sethi, D.; Duemke, J. Automated Apheresis Platforms Function as a Therapeutic and an Enabling Technology for Cell-Based Therapies. European Pharmaceutical Manufacturer. February 28, 2024. https://pharmaceuticalmanufacturer.media/pharmaceutical-industry-insights/latest-pharmaceutical-manufacturing-industry-insights/automated-apheresis-platforms-function-as-a-therapeutic-and-/
3. Safford, M.; Jerbi, T.; Koudji, M.; et al. Machine Learning Enables High-Density T-cell Expansion with Lower Costs While Maintaining Product Quality. Cytotherapy. 2024. 26 (6), S26. https://www.isct-cytotherapy.org/article/S1465-3249(24)00130-0/abstract
4. 908 Devices and Terumo Blood and Cell Technologies Collaborate to Add Online Analytics to Quantum Flex Cell Expansion System. 908 Devices. November 6, 2023. https://908devices.com/news/908-devices-and-terumo-blood-and-cell-technologies-collaborate-to-add-on-line-analytics-to-quantum-flex-cell-expansion-system/
