Digital twin-assisted process design for NK-cell therapies

Natural killer (NK) cells, that are part of the human immune system, are produced and used as Advanced Therapy Medicinal Products (ATMPs) to cure diseases for which only inadequate other therapies are available. These cells function as killer cells by inducing specifically the death of the malignant cells, called cytotoxicity. To use NK cells as a therapy, they are separated from blood, modified and cultured outside of the body to expand in number, then administered into the patient to fight against the disease. To culture these cells, their natural environment, the human body, should be mimicked to ensure therapeutic safety and effectiveness. However, the cellular attributes vary from patient to patient, therefore the parameters of the culturing process should be adjusted to the optimum of the donors NK cells. To create an optimal, adjustable culture system, the relevant parameters, that have an influence on the NK cultures performance, have to be determined and their interdependence should be explored. The aim of this project is to find the relevant parameters for NK cultures that have an impact on the production and functionality, such as cytotoxicity, of the cells. Secondly, linking process parameters to the outcome of the culturing process is even more important due to the many interlinks and co-effects between these parameters and the general complexity of these cell cultures. Therefore, we aim to use mathematical models and other statistical methods that can describe the complexity of the process. This model should be the digital twin of the culturing process and we aim to deliver it as a methodological and knowledge management platform for ATMP manufacturing. To achieve the project goals, cells will be cultured and analyzed in Germany at our project partners TUHH and UKE. The plan is that modelling tools will be used from project start, based on preliminary results, to reduce experimental effort. The cytotoxic potency and other quality-related aspects of the cells will be verified by a series of biological tests that measure cell surface markers, receptors, binding ligands. We plan to use advanced analytical techniques during the culture process as well, such as flow cytometry or spectroscopic measurements, and we will relate the analytics to cell functionality using our model. The Austrian part of the project will be the statistical evaluation of the results and the generation of the digital twin to link the cultivation process with cell functionality. This will be an iterative approach, with initial versions of the digital twin to be tested and further tuned until it is ready to be applied on a different NK cell line to demonstrate its predictive ability and superior process control. Our target is that our contribution to the field of ATMPs will become a new methodological approach that can lead to process understanding more efficiently than the conventional methods.

The DigiNK project brought together TU Hamburg, University Hospital Brandenburg, and TU Wien to advance natural killer (NK) cell therapy through a combined experimental and data-driven approach. Using primary human NK cells from a healthy donor and a breast cancer patient, the consortium explored strategies to improve both cell expansion and cytotoxicity. TU Hamburg conducted experimental campaigns, Brandenburg provided analytical measurements, and TU Wien contributed advanced data analytics, predictive modeling, and additional laboratory work. The project established a comprehensive framework based on multivariate experimental design, statistical modeling, and digital twin development. An initial Design of Experiment evaluated nutrient effects and culture duration on NK cell quality attributes, while advanced analyses such as PCA and Kalman filtering clarified growth dynamics and handling effects. Predictive partial least squares regression models were developed to link process parameters with outcomes, enabling the creation of a digital twin capable of forecasting optimal conditions across different donor sources. A subsequent targeted experiment validated and refined these models, resulting in precise nutrient feeding strategies that support scalable and automated NK cell production. TU Wien extended the project by optimizing stirred-tank bioreactor processes, developing a serum-reduced medium, and implementing a repetitive batch strategy, with results submitted for publication. The team also emphasized the need for harmonized quality standards in NK cell therapy. Key discoveries included the reversible nature of NK-92 cytotoxicity, forming the basis of a patented two-phase cultivation process, and the identification of lactate-induced FasL loss as a driver of reduced cytotoxicity in tumor-like environments, also protected by a patent application. Overall, DigiNK delivered both mechanistic insights and technological innovations. By integrating experimental biology with predictive modeling, it laid the foundation for robust digital twins, improved manufacturing strategies, and novel therapeutic concepts in NK cell therapy, with publications and patents already advancing its impact.