| Optimal experimental design is a mathematical framework, based on statistical principles, for strategically choosing experimental conditions to achieve a defined experimental objective. The experimental objective may be, for example, to produce large quantities of a chemical, or to obtain data for calibrating a predictive mathematical model of a physical system. The framework of optimal experimental design can, in principle, be applied to any experimental system.;Tangentially similar to factorial design of experiments, the aim of optimal experimental design is to maximize efficiency of experiments by (1) generating data that produce maximal information and (2) minimizing the resources used for running the experiments, including time and laboratory materials. Unlike factorial design of experiments, however, optimal experimental design involves the formulation and solution of an optimization problem, enabling the search of a continuous experimental variable space rather than choosing discrete factor levels. The optimization problem is formulated by defining the objective of the experiment, defining the experimental variables that can be controlled and optimized to achieve the objective, and by representing the experimental system with a mathematical model that relates the experimental objective to the experimentally controlled variables.;In this work, three optimal experimental design problems were formulated and solved. In one problem, the production of flavonoids, plant metabolites with potential therapeutic properties, was maximized in metabolically engineered bacteria. This led to a 65% increase in production compared to the production levels achieved before optimization. In the second problem, parameter accuracy was maximized for a mathematical model of interleukin-6 (IL-6) extracellular inflammation signaling. A substantial increase in parameter accuracy was observed when the parameters were fitted using data from simulated experiments in which the optimal designs were implemented. In the third problem, in silico differentiation of regulatory T cells (Tregs) from naive T cells was maximized relative to differentiation of pro-inflammatory T-helper-17 (Th17) cells, for use in cell-based treatment of chronic inflammation. It was shown here that time-dependent concentrations of extracellular cytokines can slightly improve Treg induction relative to Th17 induction in silico, compared to constant extracellular cytokine concentrations, but that constant cytokine concentrations may suffice experimentally. This result can potentially facilitate the induction of Tregs for clinical use to treat chronic inflammation.;The formulation and solution of these problems demonstrate the implementation of optimal experimental design for practical biological systems. The methods presented can, however, be adapted for practical experimental systems in other scientific or engineering disciplines. As optimal experimental design has traditionally, and for the most part, been applied toward proof-of-principle studies with less practical objectives, the work presented here has implications towards the further development of optimal experimental design such that it can be utilized for its actual purpose of maximizing efficiency for practical experiments conducted to solve today's scientific and engineering problems. |
