Monoclonal antibodies are nothing short of revolutionary in the field of modern medicine. These laboratory-engineered proteins play a crucial role in treating a range of conditions, from cancers to autoimmune disorders. With projections indicating that the market for these therapeutic solutions could potentially double by 2030, one significant challenge remains: the speed at which they can be produced.
However, exciting developments from the University of Oklahoma are poised to tackle this issue head-on.
In a study published in the journal Communications Engineering, researchers Chongle Pan, a professor specializing in computer science and biomedical engineering, along with doctoral student Penghua Wang, introduced an innovative machine learning model designed to significantly hasten the production timeline of monoclonal antibodies.
"We're addressing a fundamental bottleneck in the biomanufacturing process," Wang explained. "The goal is to facilitate a quicker transition to market."
In humans, antibodies are generated by white blood cells known as B cells. In the realm of biomanufacturing, however, the task is performed by Chinese hamster ovary (CHO) cells, which are the gold standard for generating therapeutic antibodies.
Pan drew an interesting analogy between this process and beer brewing. Just as yeast ferments sugars to produce alcohol, CHO cells utilize nutrients specifically formulated to maximize antibody output.
Yet, not all cloned cell lines yield antibodies with equal efficiency; their productivity can vary significantly. Therefore, biomanufacturers must meticulously screen these cultured samples—a phase that can extend over several weeks. Streamlining this process is essential for pharmaceutical companies, as it could lead to reduced medication costs for patients.
Pan and Wang proposed that it might be possible to forecast cell productivity using growth data collected in earlier stages of production. To validate their hypothesis, they collaborated with Wheeler Bio, a contract development and manufacturing organization based in Oklahoma City that specializes in antibody therapies. Wheeler provided vital production data, which the researchers integrated with a well-established mathematical framework known as the Luedeking-Piret model, which describes cellular growth and protein production, to develop and refine their machine learning algorithms.
Following extensive testing and optimization, the researchers found that their model accurately identified high-performing clones in 76.2% of instances and effectively predicted daily production rates from day 10 to day 16, relying solely on data gathered during the initial nine days of growth. The outcomes, as noted by Pan and Wang, underscore the potential for a "simulated real-world clone selection" that would allow for faster and more confident identification of high-yielding clones.
Although further tests and model adjustments are necessary before integrating their findings into Wheeler's production methods, company representatives expressed optimism about the preliminary results.
"Wheeler Bio is dedicated to harnessing artificial intelligence and machine learning to enhance our approach to cell line and process development for antibody production," stated Patrick Lucy, the President and CEO of Wheeler Bio.
He added, "This foundational work marks the initial phase of our strategic commitment to employ AI and machine learning techniques, further advancing our ModularCMC™ platform."
This research initiative was part of a $35 million program backed by the U.S. Economic Development Administration, aimed at bolstering the biotechnology sector in the Oklahoma City area. Institutions like OU's Gallogly College of Engineering and the newly launched OU Bioprocessing Core Facility play pivotal roles in this effort, which seeks to merge academic innovation with practical industrial applications.
"In academia, we often focus on theoretical studies," Pan commented. "However, this project, coupled with our collaboration with Wheeler Bio, allowed us to apply our expertise in machine learning and data science to a pressing real-world challenge faced by the industry."
About the Project
This research received support from the U.S. Department of Commerce, Economic Development Administration, under Award No. 087905677. The full study titled "Luedeking-Piret regression for multi-step-ahead forecasting and clone selection in monoclonal antibodies biomanufacturing" is available at https://doi.org/10.1038/s44172-025-00547-7
About Wheeler Bio
Wheeler Bio stands at the forefront of contract development and manufacturing, having developed the ModularCMC™ platform that expedites the transition of antibody-based therapeutics from discovery phases to clinical trials while ensuring scalability for advanced development and commercialization. This streamlined approach facilitates the journey from drug discovery to clinical manufacturing through well-defined systematic processes, ultimately delivering cGMP products supported by a comprehensive Common Technical Document (CTD) Module 3 for Investigational New Drug Applications (INDs). Wheeler's dedication to a High Science/High Touch methodology combines cutting-edge development with cGMP manufacturing technologies, backed by a highly experienced scientific team focused on transparency, scientific integrity, and genuine collaboration. The mission of Wheeler Bio is to accelerate the conversion of drug discoveries into meaningful clinical interventions for their partners and the patients they aim to benefit.
About the University of Oklahoma
/Public Release. This information provided by the originating organization or authors may be subject to change and has been edited for clarity, style, and length. Mirage.News does not endorse any institutional stances, and all opinions, positions, and conclusions expressed herein are solely those of the authors. Read the complete article here: https://www.miragenews.com/ou-industry-team-use-ai-to-speed-antibody-drug-1616382/.
