Development of a Combinatorial Search Algorithm for Optimizing Stem Cell Culture Conditions

The ability to accurately manipulate stem and progenitor (?precursor?) cell cultures has become critical with the increasing interest of novel cell-based therapies. Efficiently optimizing cell culture systems is challenging due to the complexity generated by numerous factors in various dose combinations, non-linear cell response, interacting factors, and donor-to-donor variability. The conventional experimental approach to identify an optimal condition is by factorial design which is costly and unrealistic as the number of experiments (N) increases exponentially (N=k^n), dependent on the number of factors (n) and dose levels (k). A strategy to efficiently identify the optimal conditions for precursor cell expansion would facilitate the production of precursor cell populations and the discovery of potentially more effective complex conditions. This study proposes the development of a combinatorial search algorithm to optimize culture conditions for precursor cell expansion based on direct measurement of cellular response. Preliminary algorithms based on the combinatorial approach integrating aspects from conventional search methodologies have been developed and assessed using simulated hematopoietic stem cell culture data inspired by experimental findings, consisting of 5 factors over 5 dose levels including an inhibitory factor and a pair of interacting factors. A deterministic algorithm required testing of a similar number (~33) of combinations as in a central composite design experiment but interexperimental controls resulted in additional experimental cost, reducing efficiency. A purely stochastic algorithm struggled to recognize and respond to effects of interacting factors and was unable to converge to an optimal condition. In both cases, increasing variability simulated in the data posed a challenge to an efficient optimization process. To address these limitations, a differential evolution-based algorithm is being developed to enable the algorithm to accommodate the high degree of complexity of an expansion culture system resulting from factor interactions and the non-linearity in factor effects. The algorithm will aim to address the interexperimental variability associated with cell culture data as part of the decision-making process of the algorithm. Such algorithm will provide a robust optimization process that efficiently identifies the optimal culture condition of precursor expansion culture systems, facilitating the development and translation of precursor cell-based applications in research and industry.