Development of an Evolutionary Algorithm-Guided, High-Dimensional Optimization for Discovery of Serum-Free Media Formulations for Human Primary Cell Expansion

Substitution of serum supplements with chemically-defined components and creating defined media formulations optimized for cell culture is required for controlling the quality and consistency of cell therapy products. This result in a complex, large-scale optimization problem due to potentially uncharacterized factor effects and the large number of factors and dose levels involved, making conventional optimization methods unfeasible. In this study, a high-dimensional optimization (e.g. up to 15 factors and 6 dose levels) based on experimental feedback control provided a framework for the identification of serum-free media formulations for human primary cell expansion. The model-free approach utilized Differential Evolution principles as the foundational driving force, utilizing an iterative optimization process combining experimentation with algorithmic analysis to guide the combinatorial optimization of culture components based solely on the observed biological response. The efficiency of this strategy was first demonstrated using TF-1 cells, a serum-dependent human myeloid progenitor cell line, and the applicability to wider primary cell cultures demonstrated in the expansion of T cells isolated from 3 different subjects. In both cell types, a set of serum-free formulations that supported cell expansion comparable or superior to the commonly used serum-containing formulations was observed while testing less than 1x10-5 % of the total search space in vitro. Post hoc multivariable analyses with the tested formulations and corresponding cell expansions were used to gain further insights into the contributions of individual factors and factor interactions. This study presented an approach that was capable of overcoming the challenges of high-dimensional optimization problems by demonstrating the discovery and optimization of serum-free culture formulations for human hematopoietic cell expansion culture. Conducting the entire optimization based on information generated through selective exploration of a defined solution space guided by an algorithm, and without the assistance of any pre-existing model, presents a framework to streamline and accelerate the optimization process for efficient development of optimized media formulations.Ph.D